CN101621607A - Method for eliminating image shades of digital camera - Google Patents

Abstract

The invention relates to a method for eliminating image shades of a digital camera, comprising two aspects of shade extraction and shade comprehension for processing the image shades. Obvious or even sharp brightness change exists in a boundary between a shade region and an illumination region of an image with shades, and a double-peak shade detection method is based on the characteristic to detect and process the image shades. The invention adopts a linear grey mapping method to adjust the grey of the shade region so that the grey level of the shade region is close to the grey level of the illumination region, thereby further achieving the aim of comprehending the image shades. The method can be applied to select the optimal moment to shoot the best momentary view. The invention has simple and convenient operation and remarkable effect.

Description

The method of eliminating image shades of digital camera

Technical field

The present invention relates to a kind of digital camera application formation method, particularly a kind of method of eliminating image shades of digital camera.

Background technology

At present, people have more and more higher requirement for the raising of image resolution ratio, yet shade appears in irradiation of sunlight unavoidably, image covers and the dead angle of photographing.Thereby destroyed the aesthetic feeling of image, even the situation that is difficult to recognize can appear in the content of dash area.Shade makes the information of the target subject that image top shadow zone is reflected lose or be interfered.How to eliminate the shade of image, especially, have more and more important practical sense in these epoch of pursuing high-definition image.

Shade is blocked and forms sunlight by overhead object, and it is on the low side to show as the DN value on view data.The shadow region things shows that amount of information is less relatively on the image, is difficult to interpretation; In daily shooting, make the user can't obtain desirable image, or make whole image owing to the shade of several key places has been lost the meaning; In image processing and engineering application, shade influences the operation process, even produces error result.And use the shadow removal method of high-precision shade monitoring method and use, and can true reappearance shadow region characters of ground object, increase image amount of information, improve the quality of data.So just eliminated the adverse effect of shade, thereby reached the purpose of correct recognition objective target identification.

Shadow removal generally comprises the content of two aspects: the detection of shadow region and the removal of shadow region.The shadow Detection technology can be divided into two classes, based on the method for model with based on the method for shade attribute.Need priori based on the method for model, be commonly used to handle specific scene, have bigger limitation about scene, target and light conditions.Method based on attribute then is to utilize the spectrum of shadow region and geometry character to detect shade, be the brightness character lower of utilizing image top shadow zone the earliest than surrounding pixel brightness value, if grey level histogram becomes bimodal or multimodal distributes, then adopt threshold method to carry out shadow Detection, the result of Chu Liing is that water body, low-light level atural object etc. are taken as shade like this, and shadow region high brightness atural object is taken as non-shade, and obviously error is bigger, on a large scale, complex-terrain atural object image is inapplicable.Tsait proposed a kind of shadow detection method based on constant color spaces such as HSI, HSV, YCbCr, but this method is the shadow region with navy blue and bottle green atural object flase drop easily by analyzing the brightness and the tone attribute in colored aerial image top shadow zone.

The brightness of each pixel or color all are the compound functions of solar irradiation function and clutter reflections function on the image, because this compound function is very complicated, theoretically, the true colours of eliminate shade on the image fully, recovering atural object in the shadow region almost are impossible.Traditional image enchancing method as histogram equalization, homomorphic filtering, normalized etc., all has certain effect to the shade that improves image, but the image shades after handling is still obvious.Simultaneously, when the key of problem is that these methods compensate shadow information, can not accomplish not change fully the information of non-hatched area.

In view of the foregoing, providing a kind of method of reasonable in design, obvious results eliminating image shades of digital camera, is one of these those skilled in the art problem that should address.

Summary of the invention

Purpose of the present invention is exactly in order to overcome weak point of the prior art, and a kind of method of reasonable in design, easy and simple to handle, obvious results eliminating image shades of digital camera is provided for people.

The technical solution adopted in the present invention is for achieving the above object: a kind of method of eliminating image shades of digital camera is characterized in that implementation step is as follows:

At first, coloured image is transformed to respectively on three chrominance spaces of RGB; As transform to the gray-scale map in R space, then make G=0, B=0; In like manner can be at the gray-scale map on R, G, three chrominance spaces of B;

Secondly, respectively the image on three chrominance spaces is carried out preliminary treatment, i.e. the detection of shade, promptly obtain the grey level histogram of image earlier, use M=d+ (D-d)/5 then, wherein d is a gray value less in the peak value, D is a gray value bigger in the peak value, M gained threshold value; The pixel that brightness in the image is lower than the M value is elected as a set, thereby realizes cutting apart of shadow region;

Once more, pretreated image is carried out the compensation of shade by the method for piecewise linear maps, the piecewise linear transform formula is as follows:

$g(x,y)=\left\{\begin{array}{cc}{\mathrm{\&gamma;}}_{1}f(x,y)+b_1,& 0\&le;f(x,y)\&le;{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1\\ {\mathrm{\&gamma;}}_{2}f(x,y)+b_2,& {\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1\&le;f(x,y)\&le;f_2\\ {\mathrm{\&gamma;}}_{3}f(x,y)+b_3,& f_2\&le;f(x,y)\&le;f_3\end{array}\right.$

Wherein, f (x y) is input picture, g (x y) is image after handling:

${\mathrm{\&gamma;}}_1=\frac{g_1}{{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1},$ b ₁＝0

${\mathrm{\&gamma;}}_2=\frac{g_2-g_1}{f_2-{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1},$ b ₂＝g ₁-γ ₂f

${\mathrm{\&gamma;}}_3=\frac{g_M-g_2}{f_M-f_2},$ b ₃＝g ₂-γ ₃f

At last, again with the image integration of three chrominance spaces,, obtain the coloured image after the Shadows Processing just the image overlay on three chrominance spaces.

A kind of method of eliminating image shades of digital camera is characterized in that implementation step is as follows:

At first, image is carried out preliminary treatment, i.e. the detection of shade obtains the grey level histogram of image earlier, uses M=d+ (D-d)/5 then, and wherein d is a gray value less in the peak value, and D is a gray value bigger in the peak value, M gained threshold value; The pixel that brightness in the image is lower than the M value is elected as a set, thereby realizes cutting apart of shadow region;

Then, pretreated image is carried out the compensation of shade by the method for piecewise linear maps, the piecewise linear transform formula is as follows:

$g(x,y)=\left\{\begin{array}{cc}{\mathrm{\&gamma;}}_{1}f(x,y)+b_1,& 0\&le;f(x,y)\&le;{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1\\ {\mathrm{\&gamma;}}_{2}f(x,y)+b_2,& {\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1\&le;f(x,y)\&le;f_2\\ {\mathrm{\&gamma;}}_{3}f(x,y)+b_3,& f_2\&le;f(x,y)\&le;f_3\end{array}\right.$

Wherein, f (x y) is input picture, g (x y) is image after handling:

${\mathrm{\&gamma;}}_1=\frac{g_1}{{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1},$ b ₁＝0

${\mathrm{\&gamma;}}_2=\frac{g_2-g_1}{f_2-{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1},$ b ₂＝g ₁-γ ₂f

${\mathrm{\&gamma;}}_3=\frac{g_M-g_2}{f_M-f_2},$ b ₃＝g ₂-γ ₃f

Directly detect and compensate.

The invention has the beneficial effects as follows: the present invention has adopted the method for gray scale Linear Mapping to adjust the gray scale of shadow region, makes the grey level of its gray scale and light area close, and then reaches the purpose that shade is compensated.Above method realizes simple, and operand is less, is convenient to the realization on digital camera, uses this method, can select the best moment to catch the most excellent moment; It is easy and simple to handle, effect is remarkable.To realize the requirements at the higher level of people to splendid moment.

Description of drawings

Fig. 1-the 1st, the present invention operate the black-and-white image flow chart;

Fig. 1-2 is that the present invention operates the chromatic image flow chart;

Fig. 2-the 1st has the grey level histogram of obvious double-peak feature image;

Fig. 2-the 2nd treats a image of shadow Detection;

Fig. 2-the 3rd, a figure threshold value equals 9 testing result images;

Fig. 2-the 4th, a figure threshold value equals 25 testing result images;

Fig. 2-the 5th, a figure threshold value equals 12 testing result images;

Fig. 2-the 6th, the histogram of a figure;

Fig. 2-the 7th has the bulk shade b image of smooth edges;

Fig. 2-the 8th, b figure shadow Detection result images;

Fig. 2-the 9th has the c image of the shade on the water surface;

Fig. 2-the 10th, c figure shadow Detection result images;

Fig. 3-the 1st, gray-scale transformation curve figure;

Fig. 3-the 2nd, d image to be compensated;

Fig. 3-the 3rd, the histogram of d figure;

Fig. 3-the 4th, the image after d figure mapping is handled.

Embodiment

Below in conjunction with accompanying drawing and preferred embodiment, to according to embodiment provided by the invention, details are as follows for feature:

Shown in Fig. 1-1,1-2, this method operating process:

Two kinds of implementations are arranged at present, and a kind of is directly black-and-white image to be handled, and another kind is earlier coloured image to be converted to black and white image, handles the process of recovering again after the processing then.

Owing to original image is transformed on the trichromatic gray space, is easy to realize.And using method all is emphatically in the detection and compensation of shade, and the present invention studies detection and the compensation technique about shade emphatically.

1, based on the detection algorithm of brightness

Fig. 2-1 is the grey level histogram that a width of cloth has the image of shade, can very clearly find out, the very bimodal curve of prominent features is arranged in this histogram, at these two peak regions, near middle bigger peak region is the main region that image information is concentrated, and also is non-hatched area part in the image.Because normal unblanketed image, its grey level histogram should be approximately smoothed curve, or monochrome information evenly distributes.So it mainly is causing of shadow region that the peak region on the limit that keeps left can be judged to be.And the fact also really so, based on the extraction of brightness, just is being based on these characteristics, is separation with gray value regional between bimodal, and the pixel that will be lower than this gray value is judged to the shadow region.

Therefore, the present invention can choose suitable threshold and cut apart with this feature as the standard of judging the shadow region between the zone at two peaks, and the part that will be lower than this value is chosen out, thereby detects the shadow region.For the ease of observing, non-hatched area is all changed into white.

According to the actual detection result, in two peak-to-peak parts, selection of threshold as much as possible little is beneficial to quite good detecting and goes out the shadow region, and can avoid flase drop to go out dark as far as possible but belong to unshaded zone.By the result, draw following this formula: M=d+ (D-d)/5, wherein d is a gray value less in the peak value, D is a gray value bigger in the peak value, M gained threshold value.Below be result of calculation to employed image among this chapter.

Table 2-1 selection of threshold statistical form

This table reflects the result, and this formula can adapt to the atural object complexity, and atural object is simple, even has the shadow region of the water surface; And result more satisfactory (result is referring to Fig. 2-5, figure-8 and Fig. 2-10).

As can be seen so long as bimodal the threshold value of choosing, all very similar to the result that shade extracts, the optional scope of numerical value is bigger, and the influence that brings is very little, is easy to realize by Fig. 2-3, Fig. 2-4 and Fig. 2-5.For the shade in the image with similar double-peak feature (as Fig. 2-6), all can extract effectively based on the detection algorithm of brightness.If but shade is at the low-light level radiation area, then easily and the atural object in cast shadow district obscure mutually, this brings very big trouble for the automatic extraction of shadow information.

This method in the histogram the higher area information of the gray value on bimodal right side as the Background of image.Peak gray value very low area information in left side is as prior image frame.Be fit to the shadow region area smaller or equal to the treatment of picture of area of illumination area.

Can see bimodal method,, just shade and area of illumination can be cut apart that for most pictures, the part that this method extracts nearly all is the shadow region, only has the darker spot of a spot of color when we have chosen proper threshold value.Very little and distribution irregular blotches may be a tree shade in the image, may be the roof of the darker vehicle of color, vegetation, roof etc.But be judged to shade easily and detecting for the lower material object of brightness by mistake.Owing to reasons such as self colourities, even in the sun, brightness is still very low such as, trees.Same situation also may be the very dark roof of color, and the color that these are deep in high-resolution remote sensing image, is very easy to obscure mutually with the shade of image self, therefore when detecting with bimodal method, causes a lot of inconvenience.Therefore the essential attribute of shade is exactly that brightness is very low, when the low part of brightness value in the image that color takes an advanced study having occurred very much, causes flase drop easily.At present also ripe without comparison method can detect the Zone Full of shade accurately and flase drop not take place, and bimodal method is simple and practical more desirable.

2, the shadow compensation of gray scale Linear Mapping

Image enhancement processing is occupied very big ratio in Digital Image Processing, some gray level images recover mainly to take enhancements after degeneration.The method of figure image intensifying is divided into spatial domain method and frequency domain method two big classes, and it is the enhancing that directly is treated to the basis with the pixel to image that spatial domain strengthens.Spatial domain is handled and be can be represented by the formula:

g(x，y)＝T[f(x，y)](3-1)

Wherein (x y) is input picture to f, and (x y) is image after handling to g, and T is to f operation, and (x y) is the position of image slices vegetarian refreshments.The simplest form of z operation is that neighborhood is 1 * 1 yardstick (being single pixel).In this case, g only depends on f in that (T operation becomes the gray scale transformation function for x, value y), and form is

s＝T(r)(3-2)

Wherein s and r be respectively g (x, y) and f (x, (x, gray scale y) y) in the arbitrfary point.

It is that the spatial domain image is a kind of what strengthen that the grey scale linear conversion strengthens, and just adjusts the dynamic range of image gray levels by the piecewise linear transform function.By point (r ₁, s ₂) and point (r ₂, s ₂) the shape of Position Control transforming function transformation function, (r ₁, s ₂) and (r ₂, s ₂) median will produce gray scale expansion in various degree in the output image, thereby influence its contrast, to reach the purpose (as Fig. 3-1) that strengthens image.

The piecewise linear transform formula is as follows:

$g(x,y)=\left\{\begin{array}{cc}{\mathrm{\&gamma;}}_{1}f(x,y)+{\mathrm{<figure-callout\; id="1"\; label="b"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">\; <figure-callout\; id="0"\; label="b"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">b</figure-callout>\; </figure-callout>}}_1,& 0\&le;f(x,y)\&le;{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1\\ {\mathrm{\&gamma;}}_{2}f(x,y)+b_2,& {\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1\&le;f(x,y)\&le;f_2\\ {\mathrm{\&gamma;}}_{3}f(x,y)+b_3,& f_2\&le;f(x,y)\&le;f_3\end{array}\right.---(3-3)$

Wherein

${\mathrm{\&gamma;}}_1=\frac{g_1}{{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1},$ b ₁＝0 (3-4)

${\mathrm{\&gamma;}}_2=\frac{g_2-g_1}{f_2-{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A2009100698770014H2.png"\; state="\{\{state\}\}">f</figure-callout>}}_1},$ b ₂＝g ₁-γ ₂f (3-5)

${\mathrm{\&gamma;}}_3=\frac{g_M-g_2}{f_M-f_2},$ b ₃＝g ₂-γ ₃f (3-6)

Because the influence that shade causes image mainly is that this regional brightness value is significantly reduced, this method is carried out linearity adjustment directly at this problem to the gray scale in the image.Thereby make shade obtain compensation to a certain degree.

According to above-mentioned principle, now the gray level image analysis of a panel height image in different resolution to be handled, its processing procedure is as follows.

At first open the gray-scale map (Fig. 3-2) of piece image.Distribute as can be seen from the gray value histogram (Fig. 3-3) of image, the grey value profile of this image is bimodal distribution: the gray value of its image section concentrates near the left peak, and the grey value profile at left peak is approximately 15 ± 10; The background parts of image concentrates near the right peak, and the grey value profile at right peak is approximately 85 ± 40.From the distribution of the two, two peak values are normal distribution basically, and the centre has certain gray value to intersect, and right peak scope is bigger, makes that the overall brightness of image is bright partially.

1). carry out gray scale according to the characteristics of intensity profile and adjust conversion.

The result wants to make image section to separate preferably with background parts according to the above analysis, reaches the purpose that strengthens image, can the local gray value of adjusting image.Employing is carried out conversion to the gray scale of image, makes picture contrast obtain adjusting, thereby reaches the purpose of figure image intensifying.Here the author adopts three sections linear transformation methods, regulate the gray value of intermediate interdigitated part, make bimodal on the gray value histogram separately and then regulate its gray value, optimal result is that the grey value profile with government's image becomes background gray scale and gradation of image two parts.Three sections concrete gray scales are divided into according to grey value profile: the gray scale of defeated people's image is 0～20,20～100,100～255 for three sections, the gray scale of corresponding output image is 0～100,100～180,180～255 for three sections, after the grey scale linear conversion process, the overall brightness of image has strengthened.

2). degree of comparing enhancement process.

Through after the above-mentioned grey scale linear changing image enhancement process, the grey value profile of image is partial to the histogrammic right side of gray value, dash area has weakened, also deepening of image section.Regulate the contrast of entire image again, background and picture contrast can be strengthened.To have the gray value scope now and evenly distribute, its contrast is strengthened, obtain the processing image after contrast strengthens.

3). further handle with high cap, low cap method

After brightness and contrast's adjustment, as can be seen, shade has carried out comparatively successful separating with image, but also has certain gray scale crossover phenomenon below image, should carry out respective handling again.Adopt the low cap of high cap to handle and to reach comparatively desirable effect.The low cap processing method of high cap is actually carries out certain addition and subtraction calculating to gray level image, removes some part wherein, reaches the purpose that strengthens image.

4). adjust the gray scale of image once more, finish image processing process

The left side of its gray value deflection grey level histogram of image of handling through too high cap, low cap, i.e. figure image intensifying, background has also strengthened simultaneously, and grey value profile is comparatively concentrated.Pass through the gray scale adjustment again, make its even distribution, get treatment effect to the end, as shown in Figure 3-4.

Grey scale linear changing image enhancement techniques is comparatively desirable when the processing grey value profile is the degraded image enhancing of bimodal form, and processing method easy and simple to handle, that practicality is very strong.

Above-mentioned compensation method can be fallen the influence compensation to a great extent that the script shade is caused, and let us can be seen the scenery of inside, shadow region.Grey scale linear changing image enhancement techniques is comparatively desirable when the processing grey value profile is the degraded image enhancing of bimodal form, is a kind of practical method.

Above-mentioned detailed description of the method for this eliminating image shades of digital camera being carried out with reference to embodiment is illustrative rather than determinate; Therefore in the variation and the modification that do not break away under the general plotting of the present invention, should belong within protection scope of the present invention.

Claims (2)

1, a kind of method of eliminating image shades of digital camera is characterized in that implementation step is as follows:

At first, coloured image is transformed to respectively on three chrominance spaces of RGB; As transform to the gray-scale map in R space, then make G=0, B=0; In like manner can be at the gray-scale map on R, G, three chrominance spaces of B;

Secondly, respectively the image on three chrominance spaces is carried out preliminary treatment, i.e. the detection of shade, promptly obtain the grey level histogram of image earlier, use M=d+ (D-d)/5 then, wherein d is a gray value less in the peak value, D is a gray value bigger in the peak value, M gained threshold value; The pixel that brightness in the image is lower than the M value is elected as a set, thereby realizes cutting apart of shadow region;

Once more, pretreated image is carried out the compensation of shade by the method for piecewise linear maps, the piecewise linear transform formula is as follows:

$g(x,y)=\left\{\begin{array}{c}{\mathrm{\&gamma;}}_{1}f(x,y)+b_1,0\&le;f(x,y)\&le;f_1\\ {\mathrm{\&gamma;}}_{2}f(x,y)+b_2,f_1\&le;f(x,y)\&le;f_2\\ r_{3}f(x,y)+b_3,f_2\&le;f(x,y)\&le;f_3\end{array}\right.$

Wherein, f (x y) is input picture, g (x y) is image after handling:

${\mathrm{\&gamma;}}_1=\frac{g_1}{f_1},$ b ₁＝0

${\mathrm{\&gamma;}}_2=\frac{g_2-g_1}{f_2-f_1},$ b ₂＝g ₁-γ ₂f

${\mathrm{\&gamma;}}_3=\frac{g_M-g_2}{f_M-f_2},$ b ₃＝g ₂-γ ₃f

At last, again with the image integration of three chrominance spaces,, obtain the coloured image after the Shadows Processing just the image overlay on three chrominance spaces.

2, a kind of method of eliminating image shades of digital camera is characterized in that implementation step is as follows:

At first, image is carried out preliminary treatment, i.e. the detection of shade obtains the grey level histogram of image earlier, uses M=d+ (D-d)/5 then, and wherein d is a gray value less in the peak value, and D is a gray value bigger in the peak value, M gained threshold value; The pixel that brightness in the image is lower than the M value is elected as a set, thereby realizes cutting apart of shadow region;

Then, pretreated image is carried out the compensation of shade by the method for piecewise linear maps, the piecewise linear transform formula is as follows:

$g(x,y)=\left\{\begin{array}{c}{\mathrm{\&gamma;}}_{1}f(x,y)+b_1,0\&le;f(x,y)\&le;f_1\\ {\mathrm{\&gamma;}}_{2}f(x,y)+b_2,f_1\&le;f(x,y)\&le;f_2\\ r_{3}f(x,y)+b_3,f_2\&le;f(x,y)\&le;f_3\end{array}\right.$

Wherein, f (x y) is input picture, g (x y) is image after handling:

${\mathrm{\&gamma;}}_1=\frac{g_1}{f_1},$ b ₁＝0

${\mathrm{\&gamma;}}_2=\frac{g_2-g_1}{f_2-f_1},$ b ₂＝g ₁-γ ₂f

${\mathrm{\&gamma;}}_3=\frac{g_M-g_2}{f_M-f_2},$ b ₃＝g ₂-γ ₃f

Directly detect and compensate.

Images