CN1692629A - Image processing apparatus and method - Google Patents

Image processing equipment and method

Technical field

The present invention relates to a kind of image processing equipment and method, being particularly related to a kind of wide dynamic range image transitions that will have the pixel value dynamic range wideer than conventional dynamic range that preferably is applicable to is the narrow dynamic image with narrower pixel value dynamic range, and the image processing equipment of enhancing contrast ratio and method.

Background technology

Traditionally, be widely used for such as the Image-forming instrument of video camera and stillcamera and such as the optical measuring device of the element inspection equipment that uses among the FA (factory automation) and such as the optical measuring device of the fujinon electronic video endoscope that uses among the ME (medical electronics) such as the solid-state imaging element of CCD (charge coupled device) and CMOS (complementary metal oxide semiconductors (CMOS)) etc.

In recent years, proposed a large amount of technology, be used to obtain compare image (hereinafter referred to as " wide DR image ") with wide pixel value dynamic range with the optics film photo that uses these solid-state imaging elements.

On the other hand, at present, such as the display device that is used to show moving image and rest image of CRT (cathode ray tube) and LCD (LCD), such as the projector equipment of projecting apparatus and the various printing device dynamic range of its supported pixel value of broadening not as yet, and narrow supported brightness colour code is only arranged.Therefore, current state is: even should successfully take wide DE image, also do not have to show as it obtains, projection or print the equipment of this image.

Therefore, demand to such technology (hereinafter referred to as " gray scale compress technique ") is arranged: utilize this technology, the dynamic range of the pixel value of wide DR image is narrowed down, perhaps in other words, compression brightness colour code makes the image (hereinafter referred to as " narrow DR image ") that produces the dynamic range that is suitable for display device etc.

Following paragraph will be explained the gray scale compress technique that proposes usually.The gray scale compress technique can be by redistributing wide DR image the gray scale of pixel value, make to be suitable for and can to realize simply by the gray scale of the narrower dynamic range of supports such as display device.

Yet, as mentioned above, simply the gray scale of the pixel value of wide DR image is redistributed equably to narrow dynamic range, only cause the brightness of integral image to change and reduce, thereby convert the very poor image of outward appearance that contrast reduces to.Usually, propose some and can suppress the gray scale compress technique of the loss of contrast.Below will explain three kinds of gray scale compress techniques that once proposed.

Can be used as the redistribution rule that the first gray scale compress technique comes illustrative technology to relate to determine gray scale adaptively based on the histogram of the brightness of the wide DR image of input (more particularly, calculating the gray scale transformation curve) based on the histogram of input picture.The first gray scale compress technique is a prerequisite with the area occupied that the main object in the image has big ratio, and be used for determining the gray scale transformation curve, make the brightness value of gray scale as much as possible being distributed to the peak value periphery in the histogram, thereby the contrast that suppresses main at least object reduces.

Yet, only be difficult in every kind of situation, all obtain satisfied result with the effort that distributes based on gray scale.Have a plurality of main objects at image, and have in the background and the sample situation (for example blue sky) than the area of broad of same brightness, described object usually can't obtain enough gray scales that is assigned to this place.

Can be used as the second gray scale compress technique comes illustrative technology to relate to the radio-frequency component of strengthening image before or after the gray scale conversion.The second gray scale compress technique is used for estimating to lose by the gray scale conversion part of the contrast of (or being considered to loss), and uses and for example be used for the fuzzy high frequency filter that hides and compensate the part of being lost.

The advantage of the second gray scale compress technique is: it produces the problem of the formation that depends on image unlike the first gray scale compress technique.Yet high frequency filter causes that the overshoot (overshoot) that contours of objects is partly located and the noise at flat place increase the weight of, and therefore are understood that always to guarantee desired image.

Can be used as the 3rd gray scale compress technique comes illustrative technology to relate to wide DR image is divided into low-frequency component image and radio-frequency component image, wherein only the low-frequency component image is subjected to suitable gray scale conversion process and reserved high-frequency component-part diagram picture is not changed, and both are produced a composograph mutually at last.

Because reserved high-frequency component-part diagram picture is not changed in the 3rd gray scale switch technology, so successfully avoided because the contrast that the gray scale conversion causes reduces.Yet, similar with the second gray scale switch technology, the 3rd gray scale switch technology still meets with such problem: partly locate overshoot in contours of objects, and the noise in the office, par increases the weight of, so also proposed a kind of by using nonlinear filter (for example median filter) to solve the method for this problem in the process that is divided into low-frequency component image and radio-frequency component image.

Sum up first to the 3rd above-mentioned gray scale compress technique below, the local relatively processing that they can be classified as by using neighbor realizes the technology (the first and second gray scale compress techniques) of gray scale compression, and uses the whole or big relatively area of image to realize the technology (the 3rd gray scale compress technique) that gray scale compresses.The factitious image that the former has caused its radio-frequency component only to be enhanced, and successfully do not obtain effective gray scale compression result fully.The latter so successfully obtained than the more natural image by the former acquisition, we can say that the gray scale compression is more effective because it can adjust relative low frequency content simultaneously with the reinforcement radio-frequency component.

Yet the problem that the latter meets with is: therefore this process need be mainly used in the mass storage of delay line or frame memory, makes it be unsuitable for hardware construction.For example, the 3rd gray scale compress technique needs spatial filter, be used for brightness is divided into a plurality of frequency contents, wherein can provide unartificial, effective gray scale to compress because have only when using with respect to the bigger spatial filter of image, so, be necessary a large amount of delay lines is integrated with circuit in order to allow that big spatial filter is installed.

Simultaneously, be used to make wide DR image to be subjected to the sample situation of the function that the gray scale compression handles for being intended to install on such as the output block of the imaging device of digital camera and digital stillcamera, because essential high speed signal is handled when guaranteeing pre-determined frame rate output image signal, so the function of handling for the gray scale compression of the digital stillcamera that for example will be integrated with hardware has big demand.Even,, pair demand that the compression of high speed gray scale is handled is arranged because be necessary that the image that will be monitored outputs to view finder so that determine the formation of image for the digital stillcamera that for example is used to take rest image.

As mentioned above, the tight demand to such gray scale compress technique is arranged: it only needs little of the memory span that consumes and light amount of calculation, allows to be easy to hardware construction, and guarantees big gray scale compression effectiveness.Yet, this gray scale compress technique is not proposed as yet.

In first to the 3rd above-mentioned gray scale compress technique, also there is other problem as described below jointly.

First problem relates to and the reinforcement radio-frequency component produces overshoot simultaneously in the brightness that contours of objects is partly located.

In order to suppress this overshoot, be necessary to use the two-dimension non linearity filter of relative large scale (20 * 20 pixels), yet, the two-dimension non linearity filter that is desirably in this size that realizes on the basis of software has caused the problem that the expense that is used to calculate will become high, and the two-dimension non linearity filter that is desirably in this size that realizes on the hardware foundation then causes that owing to delay line in a large number circuit scale will become big problem.

Second problem relates to the control of the contrast reinforcement amount of high luminance area and low-light level district medium-high frequency composition.The above-mentioned second and the 3rd gray scale compress technique something in common is that brightness is divided into low-frequency component and radio-frequency component, and realizes the gray scale compression by strengthen radio-frequency component when keeping suppressing low-frequency component relatively.

Yet, strengthening radio-frequency component causes producing brightness amplitude limit (clipping) at the periphery of acceptable high-high brightness such as display device and minimum brightness, thereby cause the image detail loss, it is suitable to make the gray scale conversion not say so, this has produced the needs to some countermeasure: by these countermeasures, the brightness amplitude limit is avoidable.

Another problem is: even under the situation that does not cause the brightness amplitude limit, too strengthen the image that contrast has caused having the contours of objects part that is strengthened artificially.

Summary of the invention

The present invention considers status and conceives, and its objective is and realize such gray scale compress technique: its needs little with the memory span that consumes and less amount of calculation, allow to be easy to hardware construction, and guarantee big gray scale compression effectiveness.

Another purpose be make might use less memory span, based on less amount of calculation and the contrast that suitably strengthens image based on simple hardware construction.

Image processing equipment of the present invention is characterised in that: it comprises the downscaled images generating apparatus, is used for generating downscaled images from input picture; The control information deriving means is used for obtaining based on downscaled images the control information of input picture; And the gray scale conversion equipment, be used to change the gray scale of input picture; Wherein, the contrast that the gray scale conversion equipment uses control information to proofread and correct input picture is as will be before the conversion gray scale and/or the processing of carrying out afterwards.

This image processing equipment can also comprise smoothing apparatus, is used for generating having pixel intensity L _cLevel and smooth after image, described pixel constitutes based on the interpolation of using the pixel that constitutes downscaled images to be calculated and the input picture of smoothing wherein can dispose the gray scale conversion equipment, makes brightness L based on the pixel of composing images _c, constitute the brightness L of the pixel of the image after level and smooth ₁And predetermined gain value g and image after generating contrast correction.

Can dispose the gray scale conversion equipment, make the brightness L that calculates the pixel that constitutes the image after the contrast correction according to following formula _u:

L _u＝g·(L _c-L ₁)+L ₁

Can dispose reduction means, make will input image be divided into a plurality of, calculate the mean value of the brightness of the pixel that belongs to single, and produce by with the piece equal number and have a downscaled images that described mean value constitutes as the pixel of pixel intensity.

Can dispose smoothing apparatus, make and accurately to make position, and use near the pixel that is present in the position of being made, thereby calculate the pixel intensity L of the image after level and smooth corresponding on the downscaled images of interpolation position (its for the locations of pixels that is interpolated) ₁

Can also dispose smoothing apparatus, feasible accurately making corresponding to the position on the downscaled images of interpolation position (it is with the locations of pixels that is interpolated), and use near be present in the position of being made 4 * 4 pixels, thereby calculate the pixel intensity L of the image after level and smooth based on the bicubic interpolation ₁

Image processing equipment of the present invention can also comprise: to analog-to-digital conversion apparatus, be used to make formation to be input to the pixel intensity L of smoothing apparatus image before _cBe subjected to number conversion; And logarithm inverse conversion device, be used to arrange the brightness of the pixel that constitutes the image after the contrast correction.

Can dispose the yield value setting device, make yield value g to be set according to following formula:

g＝1+(g ₀-1)·attn(Th ₁，Th ₂，L _c)

Can dispose the yield value setting device, make and calculate pad value attn (Th according to following formula ₁, Th ₂, L _c):

attn(Th ₁，Th ₂，L _c)＝|(L _c-Th ₁)/(Th ₂-Th ₁)| (2Th ₁-Th ₂≤L _c≤Th ₂)

attn(Th ₁，Th ₂，L _c)＝1 (L _c＜2Th ₁-Th ₂，Th ₂＜L _c)

Can dispose the gray scale conversion equipment, make the brightness L that calculates the pixel that constitutes the image after the contrast correction according to following formula _u:

L _u＝g·(L _c-L ₁)+L ₁

The first luminance threshold Th ₁May be defined as medium grey level (level), and the second luminance threshold Th ₂May be defined as maximum white level.

Can dispose the downscaled images generating apparatus, make by converting input picture to after the tone conversion image based on transfer function, size of images after contractive color is turned and changed then, generate downscaled images, and can the arrangement corrects information acquisition device, so that obtain the control information of the slope that comprises transfer function, and, can dispose the gray scale conversion equipment, the contrast of the image after making slope based on downscaled images and transfer function come correcting colour to turn to change.

Image processing equipment of the present invention can also comprise holding device, is used to keep corresponding to the downscaled images of the image of previous frame and is applied to the slope of the transfer function on the image of this previous frame.

Can dispose the downscaled images generating apparatus, make and use one or more transfer functions progressively to change the pixel value of the image of present frame, and can dispose the gray scale conversion equipment, make by based on the downscaled images that keeps by holding device and the contrast of proofreading and correct the image after the tint correction separately corresponding to the product of the slope of one or more transfer functions, generate the image after the contrast correction.

In one or more transfer functions, at least a transfer function can be configured to monotone increasing function.

Image processing equipment of the present invention can also comprise average computing device, be used to calculate the mean value of the pixel value of the image after tone is changed, and, in one or more transfer functions, can dispose at least a transfer function, make to have reciprocal proportional slope with the mean value that calculates by average computing device.

Can dispose average computing device, make the image after the tint correction is divided into a plurality of, and the pixel value by single of weighting summation average and calculated value, as mean value.

Can dispose the downscaled images generating apparatus, make the size of turning the image after changing generate first downscaled images by contractive color, and the single pixel value of first downscaled images be multiply by reciprocal proportional value with the mean value of the pixel value of first downscaled images, thereby generate second downscaled images.

Image processing equipment of the present invention can also comprise analog-to-digital conversion apparatus, is used for making the pixel value of the image of present frame to be subjected to number conversion, and logarithm inverse conversion device, is used to make the pixel value of the image after the contrast correction to be subjected to the logarithm inverse conversion.

Image processing equipment of the present invention can also comprise: the γ conversion equipment is used to make the pixel value of the image after the contrast correction to be subjected to the γ conversion; Brightness range information calculations device is used to calculate brightness range information, the distribution of the brightness composition of the image after the contrast correction after this brightness range information indication is changed through γ conversion equipment γ; And normalization (normalization) device, being used for based on brightness range information by the calculating of brightness range information calculations device, the distribution normalization of the pixel value of the image after the contrast correction after will changing through γ conversion equipment γ is to preset range.

Can dispose the brightness range calculation element, the higher limit of the brightness composition of the image after making calculating through the contrast correction after the γ conversion equipment γ conversion and lower limit are as brightness range information, and can dispose regular device, so that the pixel value of the image after the conversion contrast correction, higher limit and the lower limit with the scope of the reproducible brightness composition of reproducer of hypothesis is consistent respectively to make higher limit and the lower limit of brightness composition of the image after the contrast correction of being calculated by brightness range information calculations device.

Can dispose holding device, the feasible brightness range information that keeps by the previous frame of brightness range information calculations device calculating.

Described image can be the monochrome image that is made of the pixel with brightness composition.

Described image can be the coloured image that is made of the pixel with a plurality of color components.

Can dispose the downscaled images generating apparatus, make and generate first luminance picture that constitutes by pixel with brightness composition based on coloured image, first luminance picture is converted to luminance picture after the tone conversion, and generates the image after the tone conversion of the colour that constitutes by pixel based on the luminance picture after the tone conversion with a plurality of color components.

Can dispose the downscaled images generating apparatus, make difference between the value of the value of the single color component by calculating coloured image and brightness composition, calculate the product of the slope of this difference and transfer function then, and this product is added on the value of single color component of the luminance picture after the tone conversion, calculate the single color component of the image after the tone conversion.

Can dispose the downscaled images generating apparatus, the feasible mean value that passes through the brightness composition of calculating first luminance picture, calculate reciprocal proportional coefficient then with mean value, and on duty with this coefficient with the single color component of coloured image, calculate the single color component of the image after the tone conversion.

Can dispose the gray scale conversion equipment, make by generating second luminance picture that constitutes by pixel with brightness composition based on the image after the tone conversion of colour, the slope of downscaled images that keeps based on second luminance picture, by holding device and transfer function and the contrast of proofreading and correct the image after the tone conversion of the colour that is generated by conversion equipment then generate the image after the colored contrast correction.

Image processing equipment of the present invention can also comprise the γ conversion equipment, is used to make the pixel value of the image after the colored contrast correction to be subjected to the γ conversion; Brightness range information calculations device, be used for based on generating the 3rd luminance picture that constitutes by pixel, and be used to calculate the brightness range information of distribution of the brightness composition of indication the 3rd luminance picture with brightness composition through the image after the contrast correction of the colour after the γ conversion equipment γ conversion; And regular device, being used for based on brightness range information by the calculating of brightness range information calculations device, the distribution normalization of the pixel value of the image after the contrast correction of the colour after will changing through γ conversion equipment γ is to preset range.

Image processing method of the present invention is made of following steps: downscaled images generates step, is used for generating downscaled images from input picture; The control information obtaining step is used for obtaining based on downscaled images the control information of input picture; And the gray scale switch process, be used to change the gray scale of input picture; Wherein, the contrast that the gray scale switch process uses control information to proofread and correct input picture is as will be before the conversion of conversion gray scale and/or the processing of carrying out afterwards.

According to image processing equipment of the present invention and method, make and might generate downscaled images from input picture, obtain control information based on the downscaled images that is generated, and the gray scale of conversion input picture.In the processing of gray scale conversion, the contrast of using control information to proofread and correct input picture is as will be before the gray scale conversion and/or the processing of carrying out afterwards.

Description of drawings

Fig. 1 shows the block diagram of the example structure of digital camera according to an embodiment of the invention;

Fig. 2 shows the block diagram of first example structure of the DSP shown in Fig. 1;

Fig. 3 shows the block diagram of first example structure of the tint ramp correcting unit shown in Fig. 2;

Fig. 4 shows the figure of example tint ramp;

Fig. 5 shows the block diagram of second example structure of the tint ramp correcting unit shown in Fig. 2;

Fig. 6 shows the block diagram of the 3rd example structure of the tint ramp correcting unit shown in Fig. 2;

Fig. 7 shows the block diagram that the downscaled images shown in Fig. 2 generates the example structure of parts;

Fig. 8 shows the block diagram of the example structure of the mean value calculation parts shown in Fig. 7;

Fig. 9 shows the block diagram of the example structure of the contrast correction parts shown in Fig. 2;

Figure 10 shows the block diagram of the example structure of the interpolation parts shown in Fig. 9;

Figure 11 is the figure that is used for the processing of the interpolation parts shown in the key-drawing 9;

Figure 12 shows the block diagram that the yield value shown in Fig. 9 is provided with the example structure of parts;

Figure 13 shows the block diagram of the example structure of the contrast strengthening part shown in Fig. 9;

Figure 14 is the figure that is used for the processing of the brightness range normalization parts shown in the key-drawing 2;

Figure 15 shows the block diagram of the example structure of the brightness range information calculations parts shown in Fig. 2;

Figure 16 shows the block diagram of the example structure of the brightness range normalization parts shown in Fig. 2;

Figure 17 shows and can replace the block diagram of the scope shown in Fig. 2 for the example structure of the composite component of the part from the tint ramp correcting unit to the contrast correction parts;

Figure 18 is the flow chart that is used to explain the gray scale compression processing of being undertaken by first example structure of DSP;

Figure 19 is the flow chart that is used for explaining the processing details of the step S1 shown in Figure 18;

Figure 20 is the flow chart that is used for explaining the processing details of the step S2 shown in Figure 18;

Figure 21 shows the block diagram of second example structure of the DSP shown in Fig. 1;

Figure 22 shows the block diagram of first example structure of the tint ramp correcting unit shown in Figure 21;

Figure 23 shows the block diagram of second example structure of the tint ramp correcting unit shown in Figure 21;

Figure 24 shows the block diagram of the 3rd example structure of the tint ramp correcting unit shown in Figure 21;

Figure 25 shows the block diagram that the downscaled images shown in Figure 21 generates the example structure of parts;

Figure 26 shows the block diagram of the example structure of the contrast correction parts shown in Figure 21;

Figure 27 shows and can replace the block diagram of the scope shown in Figure 21 for the example structure of the composite component of the part from the tint ramp correcting unit to the contrast correction parts;

Figure 28 shows the block diagram of the example structure of the brightness range information calculations parts shown in Figure 21;

Figure 29 is the flow chart that is used to explain the gray scale compression processing of being undertaken by second example structure of DSP;

Figure 30 is the flow chart that is used for explaining the processing details of the step S43 shown in Figure 29;

Figure 31 is the flow chart that is used for explaining the processing details of the step S44 shown in Figure 29;

Figure 32 shows the block diagram of the example structure that has applied image processing system of the present invention;

Figure 33 is the flow chart that is used to explain the operation of the image processing system shown in Figure 32;

Figure 34 shows the block diagram of first example structure of the image processing equipment shown in Figure 32;

Figure 35 shows the block diagram of the example structure of the tint ramp correcting unit shown in Figure 34;

Figure 36 shows the figure of the example tint ramp that uses in first example structure of image processing equipment;

Figure 37 shows the block diagram that the level and smooth brightness shown in Figure 34 generates the example structure of parts;

Figure 38 shows the block diagram that the downscaled images shown in Figure 37 generates the example structure of parts;

Figure 39 shows the block diagram of the example structure of the mean value calculation parts shown in Figure 38;

Figure 40 shows the block diagram of the example structure of the interpolation parts shown in Figure 37;

Figure 41 shows the block diagram that the yield value shown in Figure 34 is provided with the example structure of parts;

Figure 42 shows the block diagram of the example structure of the contrast correction parts shown in Figure 34;

Figure 43 is used to explain that the gray scale compressed image that is undertaken by first example structure of image processing equipment generates the flow chart of handling;

Figure 44 shows the block diagram of second example structure of the image processing equipment shown in Figure 32;

Figure 45 is used to explain that the gray scale compressed image that is undertaken by second example structure of image processing equipment generates the flow chart of handling; And

Figure 46 shows the block diagram of the example structure of general purpose personal computer.

Embodiment

Explain digital camera below with reference to the accompanying drawings as one embodiment of the present of invention.

Next, will describe DSP 7, it is a key of the present invention.

Fig. 2 shows first example structure of the DSP 7 that is suitable for wide DR image, and this wide DR image is a monochrome image.The monochromatic wide DR image that below will be input to DSP 7 is called wide DR luminance picture L.The pixel value of wide DR luminance picture (being brightness value) is expressed as L (p).In this context, p is the vector or the coordinate of the location of pixels on the presentation video, for example p=(x, y).Therefore, determine to use L (p), it comprises the information of location of pixels and brightness value, distinguishes with the L that represents wide DR luminance picture.Identical content also will be applied on other image and pixel value thereof of describing after a while.

Design DSP 7 makes to be input on it according to the order of the grating brightness L (p) with wide DR luminance picture L.

Utilize DSP 7, the single frame of the wide DR image of input is calculated average information, and the average information of being calculated is used to handle the wide DR image that a frame arrives later on.

Although in order to carry out the compression of effective gray scale, be necessary to use the information of calculating based on the brightness value of the whole or wide region of image usually, in installation, produced before this information of calculating time delay with the problem of increase.Therefore, DSP 7 by select minimum may time dependent information, the average information of previous frame is used for gray scale compression to present frame.Even the feasible expansion that after installing, also might avoid memory consumption and circuit scale of this structure.

Next, the details of first example structure of DSP 7 will be described with reference to the drawings.

Fig. 4 shows the example of tint ramp, wherein on the normalized logarithmic axis, imports brightness L (p) by abscissa plots respectively in [0,1] scope, the brightness L after the ordinate plots tint ramp is proofreaied and correct _c(p).The application of the anti-sigmoid curve that increase, mild of dullness as shown in this example will can not cause very strong gray scale compression effectiveness in high luminance area and low-light level district, might obtain desirable have blind in vain (whiteout) of less degree or the tone of black blind (blackout) even make after the gray scale compression yet.Otherwise, the gray scale compression will influence the intermediate luminance district strongly, and this means that the contrast correction of describing after a while can be applied in the intermediate luminance district fully, and cause the desirable narrow DR image that in the intermediate luminance scope, also has the contrast correction of less degree.

It should be noted, usually can be by in the full luminance scope, seeking slope value, and the mean value of these values is defined as the typical value γ that typical value γ determines expression tint ramp slope.It is 0.67 typical value γ that tint ramp shown in Fig. 4 has value.

Suppose predetermined constant logL below _TBe defined as other logarithm brightness of middle grade, then the mean value of the logarithm brightness logL (p) of a frame is converted into and has and logL _TLogarithm brightness logL after the tint ramp of identical value is proofreaied and correct _c(p).

Although typical value γ calculates each frame, the average value mu that in fact is based on logarithm brightness logL (p) because of this value is calculated, so this value formerly is more or less the same between frame and the subsequent frame.Therefore, with above-mentioned downscaled images logL _C1With brightness range information [Y _d, Y _b] similar, go back design liaison value γ, proofread and correct with the tint ramp that former frame is used for present frame.Therefore, define and represent value γ is also included within the average information.

Mean value calculation parts 72-i (i=1,2 ..., N) from the logarithm brightness value logL of a frame _c(p) the logarithm brightness logL that is classified into the i piece is calculated in the inside _c(p) mean value, and it is outputed to composite component 73.Composite component 73 generates the downscaled images logL of m * n pixel _C1, and downscaled images memory 24 storages in the level of feasible back, wherein said downscaled images logL _C1Has the logarithm brightness logL that receives from average computing device 72-i respectively _c(p) mean value is as pixel value.

To be described with reference to Figure 11 interpolation position p, adjacent pixels value a[4 at this] relation of [4] and displacement dx, dy.

M * n shown in Figure 11 grid is represented the downscaled images logL of m * n pixel _C1Below hypothesis provide interpolation position p=(px, py), then corresponding to the downscaled images logL of interpolation position p _C1On position q be given q=(qx, qy)=(px/bx-0.5, py/by-0.5), wherein (bx, by)=(image logL _cHorizontal number of pixels/m, image logL _cVertical number of pixels/n).

In order to obtain neighbor, recommend to obtain the qx-2＜x＜qx+2 and the interior downscaled images logL of qy-2＜y＜qy+2 scope that fall into shown in the shade of Figure 11 corresponding to the position q periphery on the downscaled images of interpolation position p _C1Pixel.In by the zone shown in the shade, 4 * 4 positions that are marked with "+" are the locations of pixels that will obtain.(dx dy) is defined as poor with respect to the nearest pixel of lower left in displacement between neighbor and the interpolation position p.That is to say, displacement can be given (dx, dy)=(fractional part of qx, the fractional part of qy).

For example, three interpolation coefficient k of level _x[4] can use following formula (1) to calculate:

z＝|dx-i+2|

In addition, vertical three interpolation coefficient k _y[4] can use following formula (2) to calculate usually:

z＝|dy-j+2|

It should be noted,, can use except the formula (1) that illustrates above, any any computing formula (2) and calculate interpolation coefficient k three times as long as can obtain enough level and smooth interpolation _x[4] and k _y[4].

log L c 1 ( p ) = Σ i = 1 4 Σ j = 1 4 a [ i ] [ j ] · k x [ i ] · k y [ j ] . . . ( 3 )

Next will explain that yield value is provided with parts 93.As mentioned above, yield value is provided with parts 93 and is used to be provided with yield value g (p), and yield value g (p) is used for adjusting by the contrast reinforcing member 94 of back level the enhancing degree in the zone except that low frequency range.For the yield value of g (p)=1, contrast reinforcing member 94 neither strengthens and does not also suppress contrast.For the yield value of g (p)＞1, the enhancing contrast ratio corresponding to this value.For the yield value of g (p)＜1, suppress contrast corresponding to this value.

To the setting of yield value be described.The contrast of image suppresses by the gray scale compression, and wherein amount of suppression depends on the slope of tint ramp.For example, consider and realize the compression of strong lime degree colour code and use tint ramp and mean the strong inhibition contrast with little slope.On the other hand, use that to have slope be that 1 straight line means that as tint ramp image does not change, perhaps do not suppress contrast.

Logarithm brightness logL in input _c(p) in approaching rank in vain or black other another situation of level, be similar to the contrast enhancing possibility that is applied to the intermediate luminance district and undesirably cause the image detail loss owing to amplitude limit, therefore work as the logarithm brightness logL that imports _cWhen (p) more approaching white rank or black rank, the adjustment yield value makes and more approaches 1.

That is to say, suppose that the inverse of the γ of representative is 1/ γ=g ₀, use following formula (4) to come calculated gains value g (p):

g(p)＝1+(g ₀-1)×attn(p) ...(4)

Wherein, attn (p) is an attenuation coefficient, and is calculated by following formula (5):

attn(p)＝attn(logL _gray，logL _white，logL _c(p))

It should be noted, in formula (5), logL _GrayOther logarithm brightness of representative expression intermediate gray-scale level, and logL _WhiteThe logarithm brightness of the white amplitude limit rank of representative expression (maximum white rank), wherein, the both is the constant that sets in advance.

Subtracter 113 calculates logarithm brightness logL _c(p) and have other logarithm brightness of intermediate gray-scale level logL _GrayBetween poor (logL _c(p)-logL _Gray), and it is outputed to divider 115.Subtracter 114 calculates has other logarithm brightness of white limiting stage logL _WhiteWith logarithm brightness logL _GrayBetween poor (logL _White-logL _Gray), and it is outputed to divider 115.Divider 115 is with the output (logL of subtracter 113 _c(p)-logL _Gray) divided by the output (logL of subtracter 114 _White-logL _Gray), and the merchant outputed to absolute calculators 116.Absolute calculators 116 is calculated the absolute value of the output of divider 115, and it is outputed to amplitude limiter (clipper) 117.Amplitude limit is carried out in the output of 117 pairs of absolute calculators 116 of amplitude limiter, makes it to be adjusted to 1 when output above 1 the time, keeps then when this output is no more than 1 that it is constant, and the result is outputed to multiplier 118 as attn (p).

The output that multiplier 118 multiply by amplitude limiter 117 with the output of subtracter 112, and product outputed to adder 119.Adder 119 adds 1 with the output of multiplier 118, and the result is outputed to the back level as yield value g (p).

Ensuing Figure 13 shows the example structure of contrast reinforcing member 94.Subtracter 121 calculates logarithm brightness logL _c(p) and the interpolate value logL of downscaled images _C1(p) poor between, and it is outputed to multiplier 122.Multiplier 122 calculates the output of subtracter 121 and the product of yield value g (p), and it is outputed to adder 123.Adder 123 is with the interpolate value logL of downscaled images _C1(p) be added in the output of multiplier 122, and with the logarithm brightness logL after the contrast correction like this _u(p) output to the back level.

It should be noted the interpolate value logL of downscaled images below _C1(p) be based on value after the interpolation of downscaled images of m * n pixel, therefore only have the image logL before dwindling _cThe extremely low frequency composition.

That is to say the output (logL of subtracter 121 _c(p)-logL _C1(p)) be equal to by from former logarithm brightness logL _c(p) only deduct extremely low frequency and become the poor of branch acquisition.As mentioned above, the logarithm brightness logL after the contrast correction _u(p) by such acquisition: luminance signal is divided into extremely low frequency composition and other composition two classes, and in these compositions, strengthens composition except low-frequency component by multiply by yield value g (p), and use adder 123 that both are synthesized once again.

As known to top, design contrast reinforcing member 94 makes when getting rid of the extremely low frequency district, uses identical yield value g (p) to strengthen composition from the Low Medium Frequency district to high frequency region.Therefore, the logarithm brightness logL after the contrast correction _u(p) can not occur in when only strengthening high frequency region in the edge part office may be clearly local overshoot, and design it to obtain to have the image of the contrast that concerning eyes, strengthens very naturally.

Yet, the dynamic range of incident light may not depend on the object that will take in essence so bigly, and the gray scale of such image compression is handled and may be caused excessive gray scale compression, thereby brightness is limited in than in the dynamic range narrower range that can be reproduced by reproducer.

Figure 14 shows the pattern of the brightness range normalization process of being undertaken by brightness range normalization parts 30.In the line chart of this figure, the brightness Y after γ proofreaies and correct before the normalization of abscissa plots brightness range, the brightness Y after the normalization of ordinate plots brightness range _n, and gray scale transformation curve α represents to be used for brightness Y is converted to Y _nConversion table.

Because when its dynamic range of reservation is constant, the dynamic range that the output that outputs to the luminance picture of reproducer causes only poor efficiency utilization to be reproduced by reproducer, so carry out normalization subsequently, make the Luminance Distribution of the luminance picture Y before the brightness range normalization expand to dynamic range whole of reproducer.

For this purpose, at first, calculate the scope [Y of Nogata Figure 131 distribution of the preceding luminance picture Y of brightness range normalization _d, Y _b], as the brightness range information of the luminance picture Y before the brightness range normalization.Then, the brightness range [Y that drops on from reproducer is set _Nb, Y _Nc] top and the inside slightly brightness value Y of low side _NaAnd Y _Ns, and, determine gray scale transformation curve α, make the brightness { Y on the abscissa _Min, Y _d, Y _b, Y _MaxWith ordinate on brightness value { Y _Nb, Y _Na, Y _Ns, Y _NcCorresponding.

Determine gray scale transformation curve α, make brightness range [Y before the brightness range normalization _d, Y _b] be mapped to brightness range [Y than reproducer _Nb, Y _Nc] slightly narrow brightness range [Y _Na, Y _Ns] reason be: prevent brightness Y _NbAnd Y _NcSharp-pointed brightness amplitude limit on every side appears on the image.

It should be noted brightness value Y here _NaAnd Y _NsBe based on brightness value Y _NbAnd Y _NcAnd set in advance with suitable value.

Comparing unit 143-2 is relatively from the brightness Y (p) of comparing unit 143-1 and the value among the register 144-2, and when from the brightness Y (p) of comparing unit 143-1 during, use brightness Y (p) from comparing unit 143-1 to upgrade value among the register 144-2 less than the value among the register 144-2.Otherwise, when the brightness Y (p) from comparing unit 143-1 is not less than value among the register 144-2, will offer the comparing unit 143-3 in the level of back from the brightness Y (p) of comparing unit 143-1.

Identical content also will be applied to comparing unit 143-3 and thereafter, and wherein, after the brightness Y of a frame (p) input is finished, register 144-1 will have the minimum value Y of the brightness Y (p) that remains on wherein _Min, and register 144-2 to 144-k will have with the order that increases progressively and remain on wherein brightness Y (p) value, and will remain on brightness Y (p) among the register 144-k as the brightness Y of brightness range information _dOutput to the back level.

Comparing unit 146-2 is relatively from the brightness Y (p) of comparing unit 146-1 and the value among the register 147-2, and when from the brightness Y (p) of comparing unit 146-1 during, use brightness Y (p) from comparing unit 146-1 to upgrade value among the register 147-2 greater than the value among the register 147-2.Otherwise, when the brightness Y (p) from comparing unit 146-1 is not more than value among the register 147-2, will offer the comparing unit 146-3 in the level of back from the brightness Y (p) of comparing unit 146-1.

Identical content also will be applied to comparing unit 146-3 and thereafter, and wherein, after the brightness Y of a frame (p) input is finished, register 147-1 will have the maximum Y of the brightness Y (p) that remains on wherein _Max, and register 147-2 to 147-k will have with the order of successively decreasing and remain on wherein brightness Y (p), and will remain on brightness y (p) among the register 147-k as the brightness Y of brightness range information _bOutput to the back level.

[ j , k , l , m ] = [ h , h , c , d ] i < d [ h , g , c , d ] d &le; i < c [ g , f , b , c ] c &le; i < b [ f , e , a , b ] b &le; i < a [ e , e , a , b ] a &le; i - - - ( 6 )

The output Y of adder 157 _n(p) by following formula (7) expression, it indicates the line segment of the gray scale transformation curve α that differentiates based on the brightness Y (p) after the γ correction.

Y n ( p ) = k - j l - m ( Y ( p ) - m ) + j . . . ( 7 )

Contrast reinforcing member 175 is based on the interpolate value logL of yield value g (p) and downscaled images _C1(p), for the logarithm brightness logL of the present frame that receives from multiplier 170 _c(p), calculate the logarithm brightness logL of the contrast of enhancing with the composition except that low-frequency component _u(p).

In step S1, DSP 7 based on about the wide DR luminance picture of previous frame as calculated and average information (the second downscaled images logL that keeps _c(p), typical value γ, brightness range information [Y _d, Y _b]), the wide DR luminance picture L of present frame of input is converted to narrow DR luminance picture Y _nDSP 7 also calculates the average information about the wide DR luminance picture L of present frame.

In step S2, DSP 7 uses about the average information of the wide DR luminance picture L of the present frame that is calculated and upgrades average information about the wide DR luminance picture of the previous frame stored.

In step S3, DSP 7 differentiates after the wide DR luminance picture of the present frame of input whether have subsequent frame, and when judgement exists, and this process is returned step S1 and repeated thereafter process.Otherwise when there was not any subsequent frame in judgement, gray scale compression processing finished.

To come details among the interpretation procedure S1, the processing on pixel basis with reference to the flow chart of Figure 19.The processing of the single step that describes below is with respect to carrying out according to the object pixel (location of pixels p) of raster order input.

Next, Figure 21 shows the example structure of the DSP 7 that is suitable for wide DR image, and this wide DR image is a coloured image.It should be noted that the wide DR image that is input to DSP 7 according to raster order is not arranged such that its whole pixels have whole compositions of R, G and B composition separately, but be arranged such that have R, G and B composition any one.Hereinafter, the wide DR image (it is coloured image) that is input to second example structure of DSP 7 is called colored mosaic (mosaic) image of wide DR.The single pixel of the colored mosaic image of wide DR have R, G and a B composition which determine by locations of pixels.

To be expressed as L (p) according to the pixel value that raster order is input to the colored mosaic image of wide DR of DSP 7 below.

In second example structure of DSP 7, go 201 pairs of mosaic (demosaic) parts wherein the pixel value L (p) of each pixel with frame of different colours remove mosaic, so that make whole pixels all have the whole of R, G and B composition, thereby generate color signal [R (p), G (p), B (p)], and it is outputed to color balance adjustment component 202.Hereinafter, will be called wide DR coloured image by the image that constitutes from the color signal that goes 201 outputs of mosaic parts.

Color balance adjustment component 202 is adjusted each of R, G and B composition, so that make the color balance of entire image suitable, thereby generates color signal [R _b(p), G _b(p), B _b(p)].It should be noted, go mosaic parts 201 and color balance adjustment component 202 be as be installed in be furnished with single-deck type ccd image sensor digital VTR on those parts such.

γ correcting unit 209 makes the color signal [R that receives from logarithm inverse conversion parts 208 _u(p), G _u(p), B _u(p)] γ that is subjected to considering the γ characteristic of reproducer (for example display 11) proofreaies and correct, and the color signal [R after the γ that is obtained proofreaied and correct _g(p), G _g(p), B _g(p)] output to monochrome information calculating unit 210 and brightness range normalization parts 212.Monochrome information calculating unit 210 will be from [the R for a frame of γ correcting unit 209 receptions _g(p), G _g(p), B _g(p)] be converted to brightness Y (p), calculate the brightness range information of the distribution be used to indicate brightness Y (p), and make 211 preservations of brightness range information-storing device.Brightness range information described here is meant the information of the distribution of the brightness Y (p) that indicates a frame, and use approaches dark brightness Y most usually _dThe brightness Y that most approaches to become clear _bAnd be calculated as brightness range information [Y _d, Y _b].

Brightness range normalization parts 212 are based on the brightness range information [Y of the previous frame that is kept by brightness range information-storing device 211 _d, Y _b], conversion is from the color signal [R of the present frame of γ correcting unit 209 receptions _g(p), G _u(p), B _g(p)], making that its distribution can meet can be by the scope of reproducer (for example display 11) expression, and with the color signal [R that is obtained _n(p), G _n(p), B _n(p)] output to the back level as narrow DR image (it is a coloured image).

As below describing, second example structure of DSP 7 that is suitable for coloured image is except mosaic parts 201 and color balance adjustment component 202 are gone in interpolation, almost similar with first example structure that is suitable for monochrome image shown in Fig. 2, but slightly modified the internal structure of individual component, make to be suitable for coloured image.

Figure 22 shows first example structure of tint ramp correcting unit 204.In first example structure, brightness generates parts 221 by calculating the logarithm color signal [logR of input _b(p), logG _b(p), logB _b(p)] linearity and and generation logarithm brightness logL _b(p), and with it output to subtracter 222-R to 222-B and table reference part 224.

Subtracter 222-R is from logarithm color signal logR _b(p) deduct logarithm brightness logL _b(p), and with the result output to multiplier 225-R.Lut memory 223 has the typical value γ that had before remained on the slope of tint ramp wherein corresponding to LUT and indication with as shown in Figure 4 tint ramp.Table reference part 224 is used the LUT that is kept by lut memory 223 and logarithm brightness logL (p) correction is logarithm brightness logL _c(p), and with it output to adder 226-R to 226-B.

Multiplier 225-R multiply by the typical value γ that receives from lut memory 223 with the output of subtracter 222-R, and it is outputed to adder 226-R.Adder 226-R calculates output and the logarithm brightness logL of multiplier 225-R _c(p) and, and the logarithm color signal logR after the result proofreaied and correct as tint ramp _c(p) output to the back level.

It should be noted that below therefore the random component that is used to handle G and B composition will omit explanation with to be used to handle those of above-mentioned R composition similar.

It should be noted that below therefore the random component that is used to handle G and B composition will omit explanation with to be used to handle those of above-mentioned R composition similar.

It should be noted that therefore the random component that is used to handle G and B composition will omit explanation with to be used to handle those of above-mentioned R composition similar.

Ensuing Figure 26 shows the example structure of contrast correction parts 207.The brightness of contrast correction parts 25 generates the logarithm color signal [logR after parts 270 calculate the tint ramp correction of importing _c(p), logG _c(p), logB _c(p)] linearity and, thereby generate logarithm brightness logL _c(1p), output to interpolation position specifying part part 271 and with it and yield value is provided with parts 273.

Interpolation position specifying part part 271 obtains logarithm brightness logL _c(p) location of pixels p (below be also referred to as interpolation position p), and it is outputed to interpolation parts 272.Interpolation parts 272 use the second downscaled images logL of the previous frame that is kept by downscaled images memory 206 _C1, calculate pixel logL corresponding to interpolation position p by interpolation _C1(p), and with it output to subtracter 274-R to 274-B and adder 276-R to 276-B.

Subtracter 274-R is from logarithm color signal logR _c(p) deduct interpolate value logL _C1(p), and with the result output to multiplier 275-R.Multiplier 275-R multiply by yield value g (p) with the output of subtracter 274-R, and the result is outputed to adder 276-R.Adder 276-R is with interpolate value logL _C1(p) be added in the output of multiplier 275-R, and with the logarithm color signal logR after the contrast correction that is obtained _u(p) output to the back level.

It should be noted that below therefore the random component that is used to handle G and B composition will omit explanation with to be used to handle those of above-mentioned R composition similar.

The brightness of composite component 300 generates parts 301 and calculates the logarithm color signal [logR of input _b(p), logG _b(p), logB _b(p)] linearity and, thereby generate logarithm brightness logL _b(p), and with it output to subtracter 302-R to 302-B and table reference part 304.Subtracter 302-R is from logarithm color signal logR _b(p) deduct logarithm brightness logL _b(p), and with the result output to multiplier 316-R.

The lut memory 303 of composite component 300 has the typical value γ that remains on the slope of tint ramp wherein corresponding to the LUT of as shown in Figure 4 tint ramp and expression in advance ₁Table reference part 304 is proofreaied and correct from brightness based on the LUT that is kept by lut memory 303 and is generated the logarithm brightness logL (p) that parts 301 receive, thereby provides logarithm brightness logL _{C '}(p), and with it output to multiplier 305 and downscaled images generation parts 306.

Downscaled images generates parts 306 with logarithm luminance picture logL _{C '}Be divided into m * n piece, calculate the logarithm brightness logL of the pixel that belongs to single _{C '}(p) Zhi mean value, thus first downscaled images of m * n pixel generated, and make 307 preservations of the first downscaled images memory.

Mean flow rate calculating unit 308 calculates the average value mu by the pixel value of first downscaled images of the previous frame of the first downscaled images memory, 307 maintenances, and it is outputed to divider 309.Divider 309 is with predetermined constant logL _TDivided by average value mu, thereby calculate typical value γ ₂, and make γ ₂Memory 310 preservations.Multiplier 311 calculates typical value γ ₁And γ ₂Product as typical value γ (=γ ₁γ ₂), and it is outputed to yield value parts 315 and multiplier 316-R to 316-B are set.

Yield value is provided with the logarithm brightness logL of parts 315 based on the present frame that receives about the typical value γ of the previous frame that receives from multiplier 311 with from multiplier 305 _c(p), calculate the logarithm brightness logL that determines present frame _cThe yield value g (p) of contrast enhancing amount (p), and it is outputed to multiplier 319-R to 319-B.

Multiplier 316-R calculates the output of subtracter 302-R and the product of typical value γ, and it is outputed to adder 317-R.Adder 317-R calculate the output of multiplier 316-R and multiplier 305 output and, and it is outputed to subtracter 318-R.Subtracter 318-R deducts interpolate value logL from the output of adder 317-R _C1(p), and with the result output to multiplier 319-R.Multiplier 319-R multiply by yield value g (p) with the output of subtracter 318-R, and the result is outputed to adder 320-R.Adder 320-R calculates output and the interpolate value logL of multiplier 319-R _C1(p) and, and with the logarithm color signal logR after the contrast correction that is obtained _u(p) output to the back level.

It should be noted that below therefore the random component that is used to handle G and B composition will omit explanation with to be used to handle those of above-mentioned R composition similar.

Ensuing Figure 28 shows the example structure of brightness range information calculations parts 210.In brightness range information calculations parts 210, brightness generates the color signal [R after parts 331 calculate the γ correction _g(p), G _g(p), B _g(p)] linearity and, thereby generate brightness Y (p), and it outputed to select parts 332.Selecting parts 332 selects based on location of pixels p to generate the brightness Y (p) that parts 331 receive from brightness.That is to say, only the brightness value of the pixel of the pixel position that sets in advance is offered MIN classification element 333 and MAX classification element 336 in the level of back.

Configuration MIN classification element 333 makes k to the relatively combined serial arrangement of parts 334 and register 335, and makes by brightness Y (p) value of register 335-1 to 335-k with incremental order maintenance input.

For example, comparing unit 334-1 is relatively from the value among brightness Y (p) that selects parts 332 and the register 335-1, and when from the brightness Y (p) that selects parts 332 during, use from the brightness Y (p) that selects parts 332 and upgrade value among the register 335-1 less than the value among the register 335-1.Otherwise, when the value that is not less than from the brightness Y (p) that selects parts 332 among the register 335-1, will offer comparing unit 334-2 in the level of back from the brightness Y (p) that selects parts 332.

Identical content also will be applied to comparing unit 334-2 and thereafter, and wherein, after the brightness Y of a frame (p) input is finished, register 335-1 will have the maximum Y of the brightness Y (p) that remains on wherein _Min, and register 335-2 to 335-k will have with incremental order and remain on wherein brightness Y (p) value, and will remain on brightness Y (p) among the register 335-k as the brightness Y of brightness range information _dOutput to the back level.

Configuration MAX classification element 336 makes k to the relatively combined serial arrangement of parts 337 and register 338, and feasible brightness Y (p) value that is kept input by register 338-1 to 338-k with the order of successively decreasing.

For example, comparing unit 337-1 is relatively from the value among brightness Y (p) that selects parts 332 and the register 338-1, and when from the brightness Y (p) that selects parts 332 during, use from the brightness Y (p) that selects parts 332 and upgrade value among the register 338-1 greater than the value among the register 338-1.Otherwise, when the value that is not more than from the brightness Y (p) that selects parts 332 among the register 338-1, will offer comparing unit 337-2 in the level of back from the brightness Y (p) that selects parts 332.

Identical content also will be applied to comparing unit 337-2 and thereafter, and wherein, after the brightness Y of a frame (p) input is finished, register 338-1 will have the maximum Y of the brightness Y (p) that remains on wherein _Max, and register 338-2 to 338-k will have with the order of successively decreasing and remain on wherein brightness Y (p) value, and will remain on brightness Y (p) among the register 338-k as the brightness Y of brightness range information _bOutput to the back level.

Because being input to brightness Y (p) value of MIN classification element 333 and MAX classification element 336 is selected parts 332 and is selected, so select at interval and the suitable adjustment of the progression k of MIN classification element 333 and MAX classification element 336, make might obtain with a frame for example in the 1% or 0.1% corresponding brightness Y of top and bottom of whole pixels _d, Y _bValue.

Next, the general gray scale compression processing of second example structure of using the DSP 7 that has applied the composite component 300 shown in Figure 27 will be described with reference to the flow chart of Figure 29.

In step S41, DSP 7 (removing mosaic parts 201) removes mosaic to the colored mosaic image of wide DR, thereby generates wide DR coloured image, and is color signal [R (p) with its pixel value, G (p), B (p)] output to color balance adjustment component 202 with the order of grating.In step S42, DSP 7 (color balance adjustment component 202) adjusts R, G and B composition respectively, makes the color balance of entire image will become suitably, thereby generates color signal [R _b(p), G _b(p), B _b(p)].

In step S43, average information (the second downscaled images logL that DSP 7 also keeps as calculated based on the wide DR coloured image with respect to previous frame _c(p), typical value γ, brightness range information [Y _d, Y _b]), convert the color signal of wide DR coloured image L of the present frame of input to narrow DR coloured image Y _nDSP 7 also calculates the average information about the wide DR coloured image L of present frame.

In step S44, DSP 7 uses the average information with respect to the wide DR coloured image L of the present frame that is calculated, and upgrades the average information about the wide DR coloured image of the previous frame stored.

In step S45, DSP 7 differentiates after the wide DR luminance picture of the present frame of input whether have subsequent frame, and when judgement exists, and this process is returned step S41 and repeated thereafter process.Otherwise when there was not any subsequent frame in judgement, gray scale compression processing finished.

To come details among the interpretation procedure S42, the processing on pixel basis with reference to the flow chart among Figure 30.The processing of the single step that describes below is with respect to carrying out according to the object pixel (location of pixels p) of raster order input.

In step S51, color balance adjustment component 202 is with the color signal [R that generates _b(p), G _b(p), B _b(p)] output to logarithm converting member 203.In step S52, logarithm converting member 203 makes the color signal [R of input _b(p), G _b(p), B _b(p)] be subjected to number conversion, and with the logarithm color signal [logR that is obtained _b(p), logG _b(p), logB _b(p)] output to composite component 300.

In step S53, the brightness of composite component 300 generates parts 301 and calculates the logarithm color signal [logR of input _b(p), logG _b(p), logB _b(p)] linearity and, thereby generate logarithm brightness logL _b(p), and it is outputed to subtracter 302-R to 302-B, and table reference part 304.In step S54, table reference part 304 is proofreaied and correct based on the LUT that is kept by lut memory 303 and with the logarithm brightness logL (p) that imports and is logarithm brightness logL _{C '}(p), and with it output to multiplier 305 and downscaled images generation parts 306.

In step S55, downscaled images generates the logarithm brightness logL of the frame after parts 306 are proofreaied and correct based on tint ramp _{C '}(p) generate first downscaled images.Calculate typical value γ based on first downscaled images that generates here, ₂Also multiply by the typical value γ that calculates so here by first downscaled images that will generate ₂And generate the second downscaled images logL _C1

In step S56, the logarithm brightness logL of the present frame that multiplier 305 will receive from table reference part 304 _{C '}(p) multiply by by γ ₂The typical value γ of the previous frame that memory 310 keeps ₂Thereby, calculate the logarithm brightness logL after tint ramp is proofreaied and correct _c(p).

In step S57, for the R composition, calculate, to generate the logarithm color signal logR after tint ramp is proofreaied and correct by subtracter 302-R, multiplier 316-R and adder 317-R _c(p).For the G composition, calculate by subtracter 302-G, multiplier 316-G and adder 317-G, to generate the logarithm color signal logG after tint ramp is proofreaied and correct _c(p).For the B composition, calculate by subtracter 302-B, multiplier 316-B and adder 317-B, to generate the logarithm color signal logB after tint ramp is proofreaied and correct _c(p).

In step S58, yield value is provided with the logarithm brightness logL of parts 315 based on the present frame that receives about the typical value γ of the previous frame that receives from multiplier 311 with from multiplier 305 _c(p), calculate the logarithm brightness logL that determines present frame _c(p) the benefit value g (p) that contrast enhancing amount increases.In step S59, interpolation parts 314 use the second downscaled images logL of the previous frame that is kept by the second downscaled images memory 313 _C1, calculate pixel logL corresponding to interpolation position p by interpolation _C1(p).

In step S60, for the R composition, calculate, to generate the logarithm color signal logR after tint ramp is proofreaied and correct by subtracter 318-R, multiplier 319-R and adder 320-R _u(p).For the G composition, calculate by subtracter 318-G, multiplier 319-G and adder 320-G, to generate the logarithm color signal logG after tint ramp is proofreaied and correct _u(p).For the B composition, calculate by subtracter 318-B, multiplier 319-B and adder 320-B, to generate the logarithm color signal logB after tint ramp is proofreaied and correct _u(p).

In step S61, logarithm inverse conversion parts 208 make the logarithm color signal [logR after the contrast correction _u(p), logG _u(p), logB _u(p)] be subjected to the logarithm inverse conversion, thereby generate color signal [R by conventional axle expression _u(p), G _u(p), B _u(p)], and with it output to γ correcting unit 209.In step S62, the γ that γ correcting unit 209 is scheduled to proofreaies and correct, and with the color signal [R after the γ correction that is obtained _g(p), G _g(p), B _g(p)] output to monochrome information calculating unit 210 and brightness range normalization parts 212.

In step S63, the brightness of brightness range information calculations parts 210 generates the color signal [R after parts 331 are proofreaied and correct based on γ _g(p), G _g(p), B _g(p)] generate brightness Y (p).In step S64, the MIN classification element 333 of brightness range information calculations parts 210 and MAX classification element 336 are calculated brightness range information [Y based on the brightness Y (p) of a frame _d, Y _b].

In step S65, brightness range normalization parts 212 are based on the brightness range information [Y of the previous frame that is kept by brightness range information-storing device 211 _d, Y _b], will be from the color signal [R of γ correcting unit 209 receptions _g(p), G _g(p), B _g(p)] normalization, thus [R calculated _n(p), G _n(p), B _n(p)].In step S66, the color signal [R that brightness range normalization parts 212 will calculate like this _n(p), G _n(p), B _n(p)] export as the pixel value of the narrow DR coloured image after the gray scale compression.It is the ending of detailed explanation of the processing of the step S43 among Figure 29 herein.

Next, will explain the details of the processing among the step S44 among Figure 29 with reference to the flow chart among Figure 31.In step S71, downscaled images generates the logarithm brightness logL of the frame after parts 306 uses are proofreaied and correct based on tint ramp _{C '}(p) first downscaled images that generates is upgraded first downscaled images that is kept by the first downscaled images memory 307.

In step S73, the single pixel of first downscaled images that multiplier 312 will be upgraded by the processing of step S71, that keep by the first downscaled images memory 307 multiply by upgrade by the processing of step S72, by γ ₂The typical value γ that memory 310 keeps ₂Thereby, generate the second downscaled images logL _C1, and upgrade the second downscaled images logL that keeps by the first downscaled images memory 313 _C1

In step S74, [the R that brightness range information calculations parts 210 use based on a frame _g(p), G _g(p), B _g(p)] the brightness range information [Y that generates _d, Y _b], upgrade brightness range information [Y by the previous frame of brightness range information-storing device 211 maintenances _d, Y _b].It is the ending of detailed explanation of the processing of the step S44 among Figure 29 herein.

Here be the ending of detailed explanation of second example structure of DSP 7.

If the present invention is applied to the digital camera as in the present embodiment: it takes the picture of wide DR image, the compression gray scale is also exported it as the image that can show with narrow dynamic range on display, then make and might realize that the gray scale compression handles by such structure: its only have a large amount of minimizings to memory span indispensable in traditional gray scale compress technique (delay line that is used for frame memory or pixel serial data), and also make to obtain such output image: it never is inferior to the gray scale compression that realizes by the extensive filtering of common use and handles the output image that obtains.

This makes might realize the high-quality and cheap digital camera that had never been realized.

Figure 36 shows the example of tint ramp, and wherein on the interior normalized logarithmic axis of [0,1] scope, respectively by the wide DR image of abscissa plots L, ordinate plots is proofreaied and correct the image L after tint ramp is afterwards proofreaied and correct _cBrightness.Tint ramp shown in Figure 36 was not proofreaied and correct this value at the brightness value of normalized wide DR image above 0.5 o'clock, and proofreaied and correct this value at the brightness value of normalized wide DR image value less than 0.5 o'clock, made and proofreaied and correct less value by bigger correcting value.That is to say, proofread and correct, avoid the shade of the dark areas in the image of maximum when making when appearing at display 511 on.The typical value γ of expression tint ramp slope can be defined by the mean value of the slope that obtains respectively on the whole zone of brightness.For example, the typical value of the tint ramp shown in Figure 36 is γ=0.94.

Mean value calculation parts 552-i (i=1,2 ..., N) calculate image L after the tint ramp be categorized into the i piece is proofreaied and correct _cThe mean value of pixel intensity, and it is outputed to composite component 553.Composite component 553 generates the downscaled images logL of m * n pixel _C1, its mean value that has respectively the brightness that receives from mean value calculation parts 552-i is as pixel value.

It should be noted interpolation position p, adjacent brightness a[4] relation object between the amount of [4] and displacement dx, dy describe with reference to Figure 11 above being similar to those, therefore will omit this explanation.

Three interpolation coefficient k on the horizontal direction _x[4] usually by using above-mentioned formula (1) to calculate.

Three interpolation coefficient k on the vertical direction _y[4] usually by using above-mentioned formula (2) to calculate.

It should be noted,, can use except the formula (1) that illustrates above, any any computing formula (2) and calculate interpolation coefficient k three times as long as can obtain enough level and smooth interpolation _x[4] and k _y[4].

It should be noted below, smoothly pixel intensity (the interpolate value L of downscaled images of the image after _C1(p)) be based on the interpolate value of m * n pixel, and therefore only have the image L before dwindling _cThe extremely low frequency composition.

It should be noted the enforcement that can walk abreast of the processing among step S112 and the step S113.

For example, 521 employings of the tint ramp correcting unit in second example structure tint ramp as shown in Figure 4.That dullness as shown in Figure 4 increases, mild anti-S shape and color transfers the application of curve can not cause very strong gray scale compression effectiveness in high luminance area and low-light level district, also might obtain desirable white blind or black blind tone with less degree even make after the gray scale compression.Otherwise, the gray scale compression will influence the intermediate luminance district strongly, and this means that contrast correction can fully be applied to the intermediate luminance district, and cause also obtaining the image L of the gray scale compression of desirable contrast correction with less degree in the intermediate luminance district _uHere, to have value be 0.67 typical value γ to tint ramp.

It should be noted the enforcement that can walk abreast of the processing among step S123 and the step S124.

The present invention is applicable to such imaging signal processing circuit usually: it not only is built in the capture apparatus such as digital camera and digital static camera, also is built in the performance equipment such as display, printer, projecting apparatus etc.

Above-mentioned a series of processing can be carried out on hardware foundation, also can carry out on basis of software.For the situation of on basis of software, carrying out this series of processes, the program that constitutes this software is installed in from recording medium on the computer that is built in specialized hardware, perhaps, for example be installed to and can carry out on the general purpose personal computer of various functions after various programs are installed.

Should be understood that, in patent specification, the step that is used for describing the program that recording medium writes down not only is included on the seasonal effect in time series basis to be handled in sequence according to given, can also comprise and replace those those processing of always handling and processing parallel or that carry out separately on the seasonal effect in time series basis.

Should be understood that, in this manual, with reference to the full content of claims, specification, accompanying drawing and the summary of Japanese patent application 2002-003134 number (submission on January 9th, 2003) and 2003-003135 number (submission on January 9th, 2003), and it is herein incorporated.

The industry scope of application

As described above, the invention enables and might realize such gray scale compress technique: it only needs the small memory capacity and the light amount of calculation that will consume, and allows to be easy to hardware construction, and guarantees big gray scale compression effectiveness.

Also make and to use less memory span,, and suitably strengthen the contrast of image based on simple hardware construction based on less amount of calculation.

Claims

(according to the modification of the 19th of treaty)

1. image processing equipment is characterized in that comprising:

The downscaled images generating apparatus is used for generating downscaled images from input picture;

The control information deriving means is used for obtaining based on this downscaled images the control information of input picture; And

The gray scale conversion equipment is used to change the gray scale of input picture;

Wherein, the contrast that the gray scale conversion equipment uses control information to proofread and correct input picture is as will be before the conversion gray scale and/or the processing of carrying out afterwards.

Smoothing apparatus is used for generation and has pixel intensity L _cLevel and smooth after image, described pixel constitutes based on the interpolation of using the pixel that constitutes downscaled images to be calculated and the input picture of smoothing,

Wherein, the gray scale conversion equipment is based on the brightness L of the pixel of composing images _c, constitute the brightness L of the pixel of the image after level and smooth ₁, and predetermined gain value g and image after generating contrast correction.

Smoothing apparatus is used for generation and has pixel intensity L _cLevel and smooth after image, described pixel constitutes based on the interpolation of using the pixel that constitutes downscaled images to be calculated and the input picture of smoothing; And

The yield value setting device is used to be provided with and is used for the yield value g of contrast correction;

Wherein, the gray scale conversion equipment is based on the brightness L of the pixel that constitutes input picture _c, constitute the brightness L of the pixel of the image after level and smooth ₁, and predetermined gain value g and image after generating contrast correction; And

Conversion equipment is used for generating the image after tone is changed by change the brightness L of the pixel that constitutes input picture based on transfer function;

Smoothing apparatus is used for by making the brightness L of the pixel that constitutes the image after tone is changed _cSmoothing and generate image after level and smooth; And

Wherein, the contrast correction device is based on the brightness L of the pixel that constitutes the image after tone is changed _c, constitute the brightness L of the pixel of the image after level and smooth ₁, and yield value g, generate the image after the contrast correction; And

The downscaled images generating apparatus, the size by the image after input picture is converted to the image after tone is changed based on transfer function and dwindles this tone conversion generates downscaled images;

The control information deriving means obtains the control information of the slope that comprises transfer function; And

The gray scale conversion equipment, based on the slope of downscaled images and transfer function, the contrast of the image after coming correcting colour to turn to change.

Holding device is used to keep downscaled images that is generated by the downscaled images generating apparatus and the control information of being obtained by means for correcting;

Wherein, holding device keeps corresponding to the downscaled images of the image of previous frame and is applied to the slope of the transfer function on the image of previous frame, and

The gray scale converting means, based on the slope of downscaled images that all is kept at the previous frame in the holding device and transfer function, the contrast of the image after coming correcting colour to turn to change.

7. image processing method is characterized in that comprising:

Downscaled images generates step, is used for generating downscaled images from input picture;

The control information obtaining step is used for obtaining based on this downscaled images the control information of input picture; And

The gray scale switch process is used to change the gray scale of input picture;

Wherein, the contrast that the gray scale switch process uses control information to proofread and correct input picture is as will be before the conversion gray scale and/or the processing of carrying out afterwards.