KR101462440B1 - Apparatus and Method for Correcting Color - Google Patents

Abstract

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color correction apparatus and a color correction method, and a color correction apparatus according to an embodiment of the present invention is a color correction apparatus for applying a plurality of filters having different sizes to estimate a color fading locally or globally A correction coefficient calculating unit for calculating a correction coefficient of each image extracted by applying each of a plurality of filters, and a correction coefficient map by summing correction coefficients of pixels corresponding to the same position of each image A correction coefficient integrating unit, and a correction unit for applying the generated correction coefficient map to the pixels of the input image to output a corrected image.

Description

[0001] Apparatus and Method for Correcting Color [0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color correction apparatus and a color correction method, and more particularly, to a color correction apparatus and a color correction method capable of correcting a faded image on a medium such as a photograph, document, ≪ / RTI >

In general, the output media such as photographs, documents, and printed materials are subject to change in color as the characteristics of the medium change according to the respective environments when they are stored for a long time. In particular, old photos are exposed to light for long periods of time, or in the form of red or yellow when the humidity is high. Franziska and Henry analyzed this fading phenomenon through various experiments. Franziska experimented with color fading of prints on temperature and humidity. Experimental results show that the higher the temperature and humidity, the lower the density of CMY dye. Henry also printed the patches and gray grades of CMY on paper and exposed them to various light sources to see the physical changes of the dye. In the above experiments, we can see that the degree of fading depends on the color or area of the actual faded image. Also, restoration of color - faded images converted into digital images through physical modeling of color fade phenomenon can be seen to be constrained.

On the other hand, color homeostasis based light source estimation methods have been applied to correct color fading images. These methods are based on the assumption that the color fade phenomenon and the chromaticity change phenomenon by the light source are the same. In general, the light source estimation algorithms are based on the assumption that the color recognized by the human visual system is expressed as a product of the reflectance of the light source and the object, and the reflectance of the object can be obtained by estimating the color of the light source from the recognized color.

Among these methods, there are GWA (Gray World Assumption), WPA (White Patch Algorithm) and White Patch Retinex Algorithm (WPA). GWA compensates the chromaticity of the image based on the assumption that the average chromaticity value of each channel of the image is gray. WPA assumes that the RGB value of the pixel having the maximum brightness value of the image is the chromaticity value of the light source, do. Recently, Edmund proposed CGWR (Combining Gray World and White patch Algorithm). The CGWR models an equation that satisfies the assumptions of GWA and WPA simultaneously, and calculates the coefficients to correct the image. As a result, CGWR shows higher performance than GWA and WPA for color correction, and also shows the effect of contrast enhancement as a quadratic function correction function. However, since these methods are global correction functions, there is a problem that it is not possible to correct the regional color change of the color fading image.

The following three methods are more specific.

A. GWA  algorithm

GWA was proposed by Buchsbaum. It estimates the light source by assuming that a certain reference spatial spectrum average exists in the total visual field. In an execution condition, the image I (x, y) has an M × N size, where x and y represent the index of the pixel location. I _R (x, y), I _G (x, y) and I _B (x, y) represent the R, G, and B channels of the image, respectively. Furthermore, Ravg, Gavg and Bavg are calculated as shown in Equations (1) to (3).

GWM maintains the G channel unchanged, and the correction ratio for the R and B channels is defined as Equation (4).

Then, the R and G pixels can be adjusted by Equation (5) and Equation (6).

B. WPA

Under Retinex theory, it is claimed that perceived white (W) is associated with the maximum cone signals of the human visual system. Therefore, the method for adjusting the white balance must be to equalize the maximum values of the R, G, and B channels. In order to avoid disturbances caused by some bright pixels in the computation, it is possible to process the clustered pixels or to low-pass the image. The maximum values Rmax, Gmax, and Bmax are calculated from Equation (7) to Equation (9).

Like GWA, WPR keeps the B channel unchanged. This method defines the gain for the R and B channels as Equation (10).

Then, the adjusted R and B channels are given by Equation (11) and Equation (12).

C. CGWR

The CGWR method proposed by the Edmund method is based on GWA and WPA for white balance and color restoration. CGWR adopts quadratic mapping of intensities of chromaticity and keeps G channel unchanged. The calculation of the correction coefficient for the R channel can be expressed in the form of a matrix as shown in Equation (13).

Here, (u, v) are parameters for white balance. In Equation (13), each column represents GWA and WPA, respectively. Equation (13) can be analytically solved using Gaussian estimation or Cramer rule. Then, the corrected image, I _{CGWR , R} (x, y) is calculated by (μ, v) as shown in Equation (14).

The correction for the B channel can be computed in an analogous manner.

However, when an image obtained by scanning a faded photograph based on sunlight is applied to methods based on conventional light source estimation, it can be seen that most images are not corrected in the same manner as the original image. Unlike an image in which the chromaticity is changed by the influence of the light source, the degree of change varies depending on the characteristics of the dyes and paper characteristics of each patch, and a method of correcting the chromaticity uniformly changed by the light source is appropriate .

FIG. 1 shows a result obtained by applying a conventional light source estimation based method. Referring to FIG. 1, a composite method (CGWR) shows similar results as the original image. However, Of the total population.

Therefore, there is a need to estimate the chromaticity of the image through an approach other than the conventional light source estimation-based method, and correct chromaticity using the chromaticity.

An object of the present invention is to provide a color correction device and a color correction method capable of correcting a faded image of a medium such as a photograph, document, and printed matter in a scanner, a cellular phone, a camera, and a computer.

A color correction apparatus according to an exemplary embodiment of the present invention includes an image extracting unit for extracting an image for estimating color fading locally or globally of an input image by applying a plurality of filters having different sizes, A correction coefficient calculating unit for calculating a correction coefficient of each image extracted by applying each of the images, a correction coefficient integrating unit for generating a correction coefficient map by adding correction coefficients of pixels corresponding to the same positions of the respective images, To the pixels of the input image to output a corrected image.

Here, one of the plurality of filters is a filter having the same size as the input image.

The image extracting unit may extract an image having an average value of each pixel with neighboring pixels, and each filter may be matched for each pixel of the input image, and an average value .

Meanwhile, the correction coefficient integrating unit may assign a weight to the calculated correction coefficient, and may generate the correction coefficient map by adding the correction coefficients of the weighted pixels.

The correction coefficient integrating unit may assign different weights according to the size of the filter.

And the correction unit outputs the corrected image by multiplying the combined correction coefficients of the correction coefficient map with the pixels.

A color correction method according to an exemplary embodiment of the present invention includes extracting an image for estimating color fading locally or globally of an input image by applying a plurality of filters having different sizes, Generating a correction coefficient map by summing correction coefficients of pixels corresponding to the same position of each of the images, and outputting the correction coefficient map to the pixels of the input image And outputting the corrected image.

Here, one of the plurality of filters is a filter having the same size as the input image.

The step of extracting each of the images may include extracting an image having an average value of each pixel with neighboring pixels, matching each filter with each pixel of the input image, And has an average value for the pixels.

The generating of the correction coefficient map may include generating the correction coefficient map by adding a weight to the calculated correction coefficient and adding the correction coefficient of the weighted pixel.

The generating of the correction coefficient map may be performed by assigning different weights according to the size of the filter.

The outputting of the corrected image may include outputting the corrected image by multiplying the combined correction coefficients of the correction coefficient map with the pixels.

According to the embodiment of the present invention, it is possible to remove faded colors by performing image correction through a color correction technique on a scanned image, and by developing software (S / W) for faded color correction, You will be able to do your work conveniently. This will create additional revenue in the mobile and PC markets.

1 is a diagram illustrating a resultant image using a conventional light source estimation based method,
2 is a diagram showing a structure of a color correction apparatus according to an embodiment of the present invention,
3 is a diagram showing a resultant image for globally faded images and locally faded input images,
Figure 4 shows the integrated correction coefficients,
5 shows input images for an experiment,
Figure 6 shows a comparison of the resulting images of Figure 5 (a)
Figure 7 is a comparison of the resulting images of Figure 5 (b)
Figure 8 is a comparison of the resulting images of Figure 5 (c)
9 is a drawing showing the results of the preference test, and
10 is a diagram illustrating a color correction method according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

2 is a view showing the structure of a color correction apparatus according to an embodiment of the present invention.

2, the color correction apparatus 190 according to the embodiment of the present invention includes an image extracting unit 200, a correction coefficient calculating unit 210, a correction coefficient integrating unit 220 and a correcting unit 230, And may further include some or all of an input unit (not shown), a storage unit (not shown), and a control unit (not shown).

Here, to include some or all of them means that some of the components may be integrated into other components, and the like, and the description includes all of them in order to facilitate a sufficient understanding of the invention.

The image extracting unit 200 obtains an image for estimating color fading locally or globally using a multi-size average filter. In other words, the image extracting unit 200 may calculate the blurred image g _{i , k} (x, y) for obtaining a blurred color correction coefficient by applying an average filter having three multiple sizes, for example, as shown in FIG. 1 . This is expressed as in Equation (15).

Here, g _{i , k} (x, y) is the average filtered images, k is the number of scales (or times) using the scale three times, and i is the color channels such as R, G, m _k , n _k is the size of the average filters.

According to the embodiment of the present invention, the size of each filter kernel used in the image extracting unit 200 is 9 × 9, 35 × 35, and the whole image is a small region, an intermediate region, and an entire region Area.

More specifically, in an embodiment of the present invention, the color correction apparatus 190 may perform a multi-size GWA in order to correct a globally or locally faded image. This method is based on the GWA for estimating the light source since the assumption is suitable for a faded image.

및

는 전역적 동작(global operation)을 위한 보정계수들이다. 이와 같은 2개의 계수는 보정 비율로서 각각 R 및 B 채널들에 적용된다. 본 발명의 실시예에서 부분적으로 바랜 색을 보정하기 위한 다중 크기 기법은 다중 크기 보정 계수들을 계산하기 위하여 사용된다. 따라서 보정계수들을 얻기 위하여 M×N 크기를 갖는 바랜 입력 이미지는 <수학식 15>에 나타낸 대로, 먼저 평균 필터를 사용해서 흐릿한 영상을 구하는 것이다.Based on Equation (15), the

And

Are correction coefficients for the global operation. These two coefficients are applied to the R and B channels respectively as correction ratios. In an embodiment of the present invention, a multi-magnitude technique for correcting a partially faded color is used to calculate multi-magnitude correction coefficients. Thus, to obtain the correction coefficients, a fuzzy input image having an M × N size is first obtained by using an average filter to obtain a blurred image, as shown in Equation (15).

In other words, a blurred image is obtained using k different average filters for each channel of RGB. There are three filters according to the embodiment of the present invention, and each average filter size uses 9 × 9, 35 × 35, and the entire image average. In this case, the average filter of the largest size is replaced with the average value of the image, and since the average is used, the calculation amount can be reduced. The average image is calculated using the filters of different sizes in order to calculate the color correction coefficient for the area corresponding to each filter size. That is, as the size of the filter is smaller, the correction of the local fade color is considered, and the use of the average of the whole image is to calculate the global fade correction.

For example, according to an embodiment of the present invention, the image extracting unit 200 extracts an image having passed through each filter, wherein the pixels for each image have an average value. In other words, to obtain the average value of each pixel, the center of the filter is matched with the pixel to be averaged, or more precisely, with each subpixel of the R, G, and B channels. Then, the average value of the matched pixels is obtained by averaging the pixel values of the pixels corresponding to the regions where the filter and the input image overlap each other.

Meanwhile, in the embodiment of the present invention, the image extracting unit 200 has been described using an average filter having a multi-size as an example. However, since it is possible to use a Gaussian filter instead of an average filter, It is not particularly limited.

The correction-coefficient calculating unit 210 obtains, for example, a GWA-based correction coefficient for each pixel of the extracted (or calculated) image. In other words, the correction coefficient calculating unit 210 calculates the color correction coefficients? _K (x, y) and? _K (x, y) for the image that has passed through the respective average filters. This can be expressed as in Equation (16).

Here,? _K (x, y) and? _K (x, y) represent point adjustment coefficients for each scale k.

The correction coefficient integrating unit 220 may combine the correction coefficients calculated for the images passed through the respective average filters to generate a combined correction coefficient map. In this process, the correction coefficient integrating unit 220 may generate the map in such a manner that appropriate weights are given according to the size of the filter, and then the correction coefficients are added. The words, calculated for different size k α _k (x, y) and β _k (x, y) is of a multiplied by a suitable weight to the size of each filter α _sum (x, y) and β _sum (x , y) is generated. To apply the correction coefficients to the GWA, these coefficients are integrated into scales with weights as in Equation (17).

Here,? _Sum (x, y) and? _Sum (x, y) are integrated correction coefficients, and Wk represents the weight of each scale k. In Equation (17), the sum of the weights Wk is 1.

Also, when weighting is applied in the correction coefficient integrating unit 220, if the weight of the filter image of a small size is given a high weight, the local correction is good, but the whole image may change to gray. In addition, if the weight of the filter corresponding to the entire image size is given a high weight, there may be no local correction effect. Therefore, it is preferable that weights are appropriately given in consideration of this point.

The correction unit 230 applies the summed correction coefficients of the integrated correction coefficient map to the input image, and outputs the corrected image. In other words, the summed correction coefficients are multiplied by each pixel of the input image to output a corrected image whose pixel value is corrected. In this process, the corrector 230 adjusts the R and B pixels using both of _{sum sum} (x, y) and _{sum sum} (x, y) Can be expressed. In other words, by using the calculated correction coefficient map, obtaining a washed-out the color correction of the image by multiplying the R and B channel image I _'R (x, y) and _{I' B (x, y)} .

Where I is the resulting image.

Based on Equation (18) and Equation (19), it is only in the embodiment of the present invention that the correction is performed only for the R and B channels in the correction unit 230 based on the GWA .

If the storage unit stores data or information to be processed in the operation of the color corrector 190, the controller may control the overall operation of the components of the color corrector 190. However, if one or more of the components of the color correction device 190 are integrated and implemented in the form of an algorithm, the storage unit and the control unit may be omitted, 190) is not particularly limited.

As a result of the above configuration, when it is assumed that the entire image of the color-faded input image is changed to red according to the embodiment of the present invention, each pixel is multiplied by alpha _sum (x, y) to reduce the R component, This allows local correction of faded images.

FIG. 3 is a diagram showing a resultant image for globally faded images and locally faded input images.

3 (b) and 3 (d) are weighted images w1 = 0.4, w2 = 0.3 and w3 = 0.3, respectively, as the result images of FIG. 3 (a) . This shows the weights respectively assigned to the images extracted from the three filters having 9 × 9, 35 × 35, and the total image size.

It can be seen that all the faded colors are corrected for globally faded images and locally faded images in the result image. Particularly, it can be confirmed that the resultant image is corrected as shown in (d) of FIG. 3 (c) for a reddish input image in the lower right corner.

4 (a) and 4 (c) are correction coefficient maps for R channels, and FIGS. 4 (b) and 4 (d) to be.

Referring to FIG. 4, (c) and (d) of FIG. 4 show integrated correction coefficients according to Equation (17) and Equation (18) for FIG. 4 (c).

Α _sum (x, y) for the R channel has a ratio of 0.3 to 2.6, and β _sum (x, y) for the B channel has a ratio of 1 to 305. In other words, the correction ratios of the reddish areas in the bottom right of FIG. 3 (d) have a lower rate for the reduction of the reddish color. On the other hand, the correction ratios of the normal regions on the left side of Fig. 3 (d) have a ratio close to 1, which means the preservation of the blue channel.

Next, experimental results of the method proposed in the embodiment of the present invention will be described.

Fig. 5 is a view showing input images for an experiment, Fig. 6 is a drawing showing comparison of result images of Fig. 5 (a), Fig. 7 is a drawing showing comparison of result images of Fig. 8 is a view showing a comparison of the result images of FIG. 5 (c), and FIG. 9 is a diagram showing the results of the preference test.

In the embodiment of the present invention, the images were firstly compared with three faded images as shown in FIG. 5 through experiments. Then the last selected image is obtained by the observer's preferred test. The results of the method proposed in the embodiment of the present invention were compared with the results of GWA, WPR, CGWR and MSR. The proposed method showed the best color enhancement for locally faded images compared to the general method in Figs. 6-8.

In the actual evaluation, the observer's preferred test was performed including 15 observers aged 25 to 34 years. Observers were asked to select an image based on criteria for fading color between the input image and the resulting image. As shown in FIG. 9, most observers have selected the method proposed in the embodiment of the present invention for partially faded images.

As a result, embodiments of the present invention have proposed a multi-size GWA to compensate both globally and locally faded images. In order to consider locally faded images, embodiments of the present invention integrate, for example, GWA and multisize techniques. The proposed method shows color correction in both globally and locally faded images in the experiment. However, methods for finding more robust and stable weights in full automatic mode and for enhancing the local luminance contrast for faded images may be proposed.

10 is a diagram illustrating a color correction method according to an embodiment of the present invention.

Referring to FIG. 10 with reference to FIG. 2, a color correction apparatus 190 according to an exemplary embodiment of the present invention applies a multi-size filter to an image for locally or globally estimating a color of an input image (S1000). For example, the color correction device 190 may extract an image for estimating a fuzzy color by applying a plurality of filters to the input image, respectively. At this time, the multi-size filters have different sizes, And may be an average filter or a Gaussian filter having a size in the order of lower, middle, and phase. In this case, the filter having the largest size may have the same size as the input image.

In addition, the color correction device 190 calculates correction coefficients of the extracted image passing through the multi-size filter (S1010). In this respect, since it has been described with reference to Equation (13), further explanation will be omitted.

Then, the color correction device 190 generates a correction coefficient map integrated by combining the calculated correction coefficients. In this process, a weight can be additionally given (S1020). Here, the weights may be given differently depending on the size of the filter. However, it is preferable that the weight is given in a range where the sum is 1.

Then, the color correction device 190 applies the correction coefficient map integrated in one to the pixels of the input image to output a corrected image (S1030). In other words, the corrected image can be outputted by multiplying each pixel of the input image by the combined correction coefficient. For example, the color correction apparatus 190 corrects the correction coefficient calculated by adding the correction coefficients of pixels corresponding to (1,1) in each image extracted by each filter to the pixel of the pixel corresponding to (1,1) of the input image Value to correct the pixel value.

Other details related to the color correction method are described in detail with reference to FIG. 2, so that further explanation is omitted.

While the present invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not limited to the disclosed embodiments. That is, within the scope of the present invention, all of the components may be selectively coupled to one or more of them. In addition, although all of the components may be implemented as one independent hardware, some or all of the components may be selectively combined to perform a part or all of the functions in one or a plurality of hardware. As shown in FIG. The codes and code segments constituting the computer program may be easily deduced by those skilled in the art. Such a computer program may be stored in a computer readable medium and read and executed by a computer to implement an embodiment of the present invention. As the storage medium of the computer program, a magnetic recording medium, an optical recording medium, a carrier wave medium, or the like may be included.

While the invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention.

200: image extracting unit 210: correction coefficient calculating unit
220: correction coefficient integrating unit 230:

Claims (12)

An image extracting unit for extracting an image for estimating color fading locally or globally of an input image by applying a plurality of filters having different sizes;
A correction coefficient calculating unit for calculating a correction coefficient of each image extracted by applying each of the plurality of filters;
A correction coefficient integration unit for generating a correction coefficient map by adding correction coefficients of pixels corresponding to the same position of each of the images; And
And a correction unit applying the generated correction coefficient map to the pixels of the input image to output a corrected image,
Wherein the image extracting unit extracts an image having an average value of each pixel with neighboring pixels and matches each filter with each pixel of the input image to calculate an average value of pixels of the area in which the input image overlaps each filter And the color correction device. The method according to claim 1,
Wherein one of the plurality of filters is a filter having the same size as the input image. delete The method according to claim 1,
Wherein the correction coefficient integrating unit adds weight values to the calculated correction coefficients and adds the correction coefficients of the weighted pixels to generate the correction coefficient map. 5. The method of claim 4,
Wherein the correction coefficient integrating unit assigns different weights according to the size of the filter. The method according to claim 1,
Wherein the correction unit multiplies the combined correction coefficients of the correction coefficient map by the pixels, and outputs the corrected image. Extracting an image for estimating color fading locally or globally of an input image by applying a plurality of filters having different sizes;
Calculating a correction coefficient of each image extracted by applying each of the plurality of filters;
Generating a correction coefficient map by adding correction coefficients of pixels corresponding to the same position of each of the images; And
And applying the generated correction coefficient map to the pixels of the input image to output a corrected image,
The step of extracting each of the images may include:
Wherein each of the pixels has an average value with respect to pixels in an area in which the input image is superimposed by matching each filter with each pixel of the input image, . 8. The method of claim 7,
Wherein one of the plurality of filters is a filter having the same size as the input image. delete 8. The method of claim 7,
Wherein the step of generating the correction coefficient map comprises:
Wherein the correction coefficient map is generated by applying a weight to the calculated correction coefficient and summing the correction coefficients of the weighted pixels. 11. The method of claim 10,
Wherein the step of generating the correction coefficient map comprises:
Wherein different weights are assigned to the filters according to sizes of the filters. 8. The method of claim 7,
The step of outputting the corrected image may include:
Wherein the corrected image is output by multiplying the combined correction coefficients of the correction coefficient map with the pixels.

Images