TW201032602A - Method for automatic white-balance calibration - Google Patents

Abstract

A method for automatic white-balance calibration is provided. The method firstly uses a predefined gain. Each color value for pixels of an inputted image is calculated. A number of the points representing the pixels within each gray region of a color space is then obtained. Next, a new gain is obtained by linearly combining those points. Further, in order to overcome a possible color deviation on account of the overlapped color regions in the color space, the method introduces a specific deletion ratio to cancel partial points whose positions located on the overlapped regions. Therefore the method provides a more proper image. Furthermore, users may decide the level of the color deviation by adjusting the predefined gain.

Description

201032602 IV. Designated representative map: (1) The representative representative of the case is: (4). (2) Simple description of the symbol of the representative figure: S401 Input image S403 introduces a chromaticity space and preset gain value S4〇5 analyzes the image S407 calculates the drop point in the color space S4〇9 counts the number of points of the gray block

S411 uses the drop point area and relative gain to derive a new gain value. 5. If there is a chemical formula in this case, please disclose the chemical formula which best shows the characteristics of the invention. 6. Description of the invention: [Technical field of the invention] The present invention is an automatic white balance The correction method, especially by using the number of gray areas in the color space, linearly combines to obtain a new white $ gain value, and rotates a more correct image. _ [Prior Art] The human eye can automatically adjust to the light source under different light sources, so that the white objects still appear white, but the digital recording devices, such as digital cameras and digital cameras, just faithfully present the reflected light of the object. Proportion, so under the green light of 201032602, 'white objects will have green, and under yellowish street lights' will have yellow. Although the human eye does not see white objects as white when all the light sources are completely corrected, the digital recording device will cause serious chromatic aberration problems. Automatic white balance (AWB) is developed to solve this problem. . The automatic white balance that is generally developed on the basis of the Gray World assumption is that the average R (red value) is equal to the average G (green value) equal to the average B (blue value). The gain value of rGB (gain). For example, please refer to the first figure in the rg color space (c〇l〇r space diagram, r=R/(R+G+B) ' g=G/(R+G+B)) Gray or white areas of various light sources, where the color temperature is approximately 3〇0〇K to 6500K. In the conventional method of obtaining the gain value, a specific light source is first obtained, including D65, D50, CW, A, FL, etc. shown in the figure, followed by rg color space in the figure (others also have RGB, YCbCr , HIS, HSV, etc., select a possible gray area, calculate the accumulated R, G, b values in this gray area' and derive the gain value according to G/R, G/B, etc. The method of automatic white balance disclosed in the prior art is disclosed in U.S. Patent No. 7,423,779. The purpose of the automatic white balance mechanism is to determine the image by analyzing the number of white pixels in a predetermined area of a color space. Light sources, such as the G/R and g/Β values of the various elements of the image, are drawn in the color space, and there is a tendency to fall in a light source area (the largest number), thereby judging which are the white pigment and the light source, and then comparing The average value of the R, G, and B values of the white pigment to adjust the gain value for the white balance. Please refer to the schematic diagram of the prior art shown in FIG. 2 for performing automatic white balance using the average values of R, G, and B. The figure shows that a color processing unit 4 201032602 207 receives the original gain value of each color pixel value (2〇1). , 2〇3,2〇5), after the comparison of the average values of the above values, find the gain value that should be adjusted for each color, then ride the self-balancing with the adjusted gain value, and then store or display the image. Corrected R, G, B values (storage/display unit 2〇9). This system can be applied to images under a variety of light sources, and the RGB values exhibited by each source are different, so the gain variation will be different. This prior art technique is to select the area of each of the white/grey areas or the most numerous numbers, and then use this area to accumulate RGB values to calculate the gain value. However, because the color gamut of different light sources will have overlapping parts, it is still a phenomenon that is easy to cause misjudgment. For example, when shooting green grass, skin color, dusk, blue sky, etc., it will cause color shift due to judgment of other light sources. . SUMMARY OF THE INVENTION The second white balance correction method proposed by the present invention has all the color space points or gain values, and the number is combined to make a new problem, so that the output color gamut overlaps and the color gamut may be misjudged. Use the falling [: group preset gain value] and then calculate the (10) ash (four) domain m 彡 这些 in these areas, and calculate the value according to the weight and the preset value added. 2) 'G ( Green), B (blue) at the color gain value 'for automatic white balance. Some of Nie Lucheng's ^ different light sources have overlapping problems in the color gamut. In this case, the invention is connected: the falling point will cause misjudgment into other light sources to produce color, and J will adjust the placement of the overlapping areas. With appropriate weights, solve the problem of possible color cast with 5 201032602. The features of the invention include: (1) = job gain value, the (10) combination produces a new gain () value allows the user to control the desired week, ° (3) to avoid easy to judge the wrong color. First burst: Month, the preferred embodiment of the 'automatic white balance correction method includes (^2 ^ ^ through the analysis of the image to capture the "red" of each element ... / you then five if value, then calculate in a specific color space The point of falling, this is true; the part of the color space has fallen according to the point of overlap between the color gamut, find out that it may be mistaken to take the test - deduct the value, by eliminating Special 'color-shifted points to eliminate the number of proportions that may be misjudged. After the point of the 仵 仵 仵 在 point in the gamut overlap part of the point, according to the deduction ratio: value: delete part of the point of overlap The point of the region produces a new set of gain values for the purpose of performing automatic white balance. [Embodiment] The invention is also based on the general gray world (the automatic white balance of n: based on the assumption of the kiss (4)) Pre-set the value of the two-plate value 'k preset light source gain value and the preset white (gray) materiality to form a new gain value. From the general use of the color space map, you can know that 'different light source' The gamut will have overlapping parts, so it is easy to make = source misjudgment (four) questions, so that After the gain value 驴 color is produced, the present invention can correct the area where the light source 6 201032602 overlaps the misjudgment by adjusting the weight ratio, so that it is corrected in some error-prone parts. See the third A color space test chart used in an embodiment of the present invention, the color space displays a relationship between red and color ratios of each element of an input image, wherein the vertical axis is displayed as b=B/G, that is, each scene< = The ratio of the blue value of the checker to the green value, and the horizontal axis shows r==^yQ, and also the ratio of the red value of the image element to the green value. Grey$ including D65, D50, FL, CW, Α and H〇r

The (or white) area can calculate the gain value belonging to a specific light source according to different gray areas in the figure. Here, D65, D50, CW, A, Hor, FL, etc. are used to establish a preset gray area to obtain a specific gray area. The preset gain value of the light source, and the six sets of preset gain values can be obtained from the light sources shown in the figure, which can be represented by the following equations (1), (2) and (3):

PreDefGain—R[i] (red gain value, 1-6) (1)

PreDefGain—G[i] (green gain value, i=l-6)... one (7)

PreDefGain_B[i] (blue gain value, i=i_6)... One (3) If the more light sources, the larger the range of color temperatures covered, the more gain values will be obtained. After paying the preset gain value, the input image is analyzed, and the R, G, and B values in the image are obtained, and the pixei is converted to the rb color space as shown in the third figure, and then each point is analyzed. In which preset gray area. For example, 'If there is a point falling on the gray area of the D65 light source, the number of D65 gray areas is accumulated (the first accumulated value c〇unt[1]); if there is a point falling on the gray area of the D50 light source , then the number of the area is accumulated (the second accumulated value c〇unt[2]), and so on, may generate more 7 201032602 cumulative value. The above accumulated value may use the value of the number of points in all the gray areas, and may also use the number of points of the gray area in the partial color space, for example, the value of a plurality of gray areas with a larger number of accumulated. The gain value of this image can be obtained by multiplying the cumulative value of each region by the preset gain value of each region, as shown in the following equation (4):

Gain_new= (Xcount[i]*PreDefGain[i])/(Zcount[i]) .....(4) where i represents the gray area in the color space, as shown in the third figure. In the case where there are six gray areas, the range of i ranges from 1 to 6. Alternatively, only the count[i] value can be selected as the first few for linear combination, that is, the values of all gray areas can be used. A part of the number is larger; count[i] represents the cumulative value of each gray area; PreDefGain[i] is the preset gain value of each gray area, and is composed of the above equations (1), (2) and (3). Multiply the number of gray areas in the color space (such as the first accumulated value count[l]) by the preset gain value of the area (preDefGain[l]), and then accumulate the number of the second gray area (such as the second The accumulated value count[2]) is multiplied by the preset gain value of the area (PreDefGain[2]), and after the gray areas in the color space are accumulated, the number of points falling in each gray area is divided, that is, Get a new gain value. The above steps are as shown in the fourth figure, which is a flow chart of the automatic white balance correction method provided by the present invention. At the beginning of the step, the system (including the device or the circuit) for performing the automatic white balance correction method of the present invention receives an input image (step S401), and then introduces a preset color space (such as the third image) and the corresponding gray regions. The preset gain value (step S403) 'After analyzing the image, including the R (red), G (green), B (blue) values in each element in the image, and calculating the ratio between the values of 8 201032602 , such as B/G, R/G, etc. (step S405). Then, after the coordinates of each point are drawn in the color space, points falling in one or more gray areas can be obtained, and at this time, the number of points of each gray area in the color space is calculated (step S407), and each gray area is accumulated. The number 中 in the equation 'i' is the count[i] of the equation (4) (step S409), and then the new gain value is calculated using the above equation (4) (step S411). Finally, an automatic white balance program is executed in accordance with this new gain value (step S413). It is worth mentioning that if the preset preset gain value is such that the areas in the color space are standard gray, the new gain value obtained by the above equation (4) is equivalent to the conventional technique using the color values. The gain value obtained by the ratio, that is, the gain value obtained by G_sum/R_sum, G_sum/B_sum, where G_sum is the sum of the green values, R-sum is the sum of the red values, and B_sum is the blue The sum of the values. However, the present invention uses a preset gain value to determine a new gain value, which provides the user with the flexibility to determine the desired color shift. If the user sets a preset gain value to generate a specific color shift effect (such as bluish Or reddish), at this time, the gain value set by the user calculates a new gain value through equation (4), and the white balance adjusted by the gain value causes the entire image to have a color shift (including a cold color or a warm color). The effect of the hue), which gives the user more flexibility in adjusting the white balance. In addition, from the relationship between the gray (or white) regions in the color space, there is a phenomenon of partial overlap. After the above image analysis, falling on the heavy parts of these color gamuts will lead to the problem of misjudging the light source. It is also often the case that the light source is misjudged. However, the automatic white balance method disclosed in the present invention introduces the concept of weighting using each light source region, and can solve the problem of false positive judgment of the light source due to color gamut overlap. 9 201032602 The overlapping of gamuts is often the main cause of color cast. See the schematic of gamut overlap in color space as shown in Figure 5. It shows that the green area (A1) in the color space overlaps with the gray area of the FL light source area (A2). During the analysis, it is impossible to judge whether the point falling in the FL light source area (A2) is a FL light source or a green object (such as a green grass). If all the books are converted, they fall in the FL light source region (A2), and in the present invention, it is judged to be the FL light source. However, if there is a point falling in the green area (Al) but being outside the FL light source area (A2), that is, falling in the area A3, the present invention will count the number of points falling in the FL light source area (A2). The number of areas falling in the area A3 is deducted according to an appropriate ratio, that is, as in equation (5), so that some possible color shift problems can be handled. FL_count-Green_count*p% ..................(5) where FL_count is the number of points falling in the FL light source area (A2),

Green-count is the number of points falling in the green area (A3) minus the FL light source area (A2), p% is an adjustable deduction ratio, which is a value less than 100%, not 100% because still Avoid points that deduct possible light sources.

The FL light source is listed here and can also be applied to different light sources. 7 Youyuan, not limited to the adjustment of the phenomenon of color gamut overlap, the equation (c〇um[i] value of (4), conforms to one or more gray areas," into a square shape. According to this method, the same It can deal with the problem of color shift of skin color, blue sky and i color eccentricity. The sixth picture of fainting, etc. shows the flow chart of the method of correcting the white balance of the light source by considering the color gamut overlap. In step 丽, input an image into the system performing the method 201032602 of the present invention, and then introduce a color office p arch R α preset addition (step s6 〇 3), followed by 3 pairs of R (8) of each pixel in each gray region. , G (green) followed by: like: including the proportion of the operation (step, step S605), such as B / G, r / G know A + ' and according to the points between the values. t 彳纟, β then calculate the color Falling points in the space ^ After the positions of the green dots are drawn in the color chamber, the points falling in one or more gray areas can be obtained (step 8607). There is a problem that the overlap between the overlaps may result in a color shift (different from the effect of the preset color shift). These areas are appropriately weighted. For example, in step s6〇9, the present invention deducts the ratio value for deducting the point in the heavy area between the gray areas, which can be overcome by eliminating the proportion of possible misjudgments. The color shift is possible. Then, as in step S6U', the falling points in the color space are judged, and the falling point which may be misjudged is found, that is, the point where the falling point is in the color region ##, according to the deduction of the closing value, the cutting is performed. The share points are in the area, and the number of points in each gray area after the deletion is deleted. In this case, after deducting the points that may be misjudged, a new cumulative value corresponding to each gray area is generated instead of the equation. (4) of count[i] ' as in equation (6):

Gain_new = (Icount_new[i]*PreDefGain[i])/(lcount_new[i]) ... - (6) where count_new[i] is the new accumulated value after deleting the point where the partial point falls on the overlapping area. Then, using the above-mentioned equation (6)', the new gain value is obtained using the falling point area and the relative gain value (step S613), and finally the automatic white balance program is executed in accordance with the new gain value (step S615). 11 201032602 — “There is a problem of overlapping areas between the gamuts of different light sources in the color T. The falling points of the parts of the ft field will cause misjudgment to other sources and the color will be biased. The overlapping area (4) points adjust the appropriate weights to solve the problem of possible color shift. In addition to the linear profit and loss value, the present invention can avoid the error of the new one that is easy to judge. Color, and can adjust the preset dish to let the user control the color tone that you want to present. To provide an automatic white balance correction method, the main gain value, according to the predefined == two out = domain The number is used as the weight distribution 4 ° ten different from the 1 ^ area, and the error that may cause the color shift is better than the good white balance effect. However, the above is only the preferred limitation of the present invention. The scope of the patent, so the use of: 1β case, not the content of the equivalent structural changes ^ ^ Mingshu and the picture, combined with Chen Ming. 1 door is included in the scope of the invention [simple description of the chart] - Figure shows the conventional technology rg color empty Schematic diagram; the second diagram shows the schematic diagram of the system using r, g, and dynamic white balance for the prior art; and the value of the thousands of sentences is shown in the figure of the color space reference map of the present invention. Shown is the flow of the automatic white balance correction method provided by the present invention, and the fifth figure shows the schematic diagram of the color gamut overlap in the color (4); the sixth figure shows the automatic 12 201032602 white balance correction method considering the case where the color gamut overlap produces the light source misjudgment Flowchart [Description of main component symbols] Area Al, A2, A3 Color processing unit 207 Gain value 201, 203, 205 Storage/display unit 209 Green area A1 φ FL source area A2 Area A3 13

Claims (1)

201032602 VII. Patent application scope: 1. An automatic white balance correction method, comprising: inputting an image; introducing a color space and a preset gain value of a plurality of gray regions in the color space; analyzing the image 'including extracting The red, green, and blue values of each pixel; the falling points of each element in the color space of the s10 nose 3 jersey image, including judging the falling points of each gray area; Φ accumulating the falling points in each gray area The number is calculated by using the number of falling points in each gray area and the preset gain value; and finally, the automatic white balance program is executed according to the new gain value. 2. The automatic white balance correction method according to claim 1, wherein the new gain value is obtained by the following equation: (Ecount[i]*PreDefGain[i])/(Xcount[i]), wherein i represents the gray area in the color space, and count[i] represents the cumulative spring value of each gray area ' PreDefGain[i] is the preset gain value for each gray area. 3. The automatic white balance correction method according to item 2 of the application scope of Shen Qing, wherein the value of δHai juice is the cumulative value of the number of points in all gray areas, or the part with a large number of accumulated points. The value of the gray area. 4. The automatic white balance correction method according to the application scope of the patent application, wherein the 忒 automatic balance correction method is more suitable for calculating the overlap point of each element in the color space, and determining the overlapping area between the gray areas. point. 5. The automatic white balance correction method according to item 4 of the patent application scope, wherein the 3H automatic balance correction method further introduces a deduction ratio value to delete the point where the resection point is at the overlapping area. 6. The automatic white balance correction method according to claim 5, wherein the new gain value is obtained by the following equation: (Ecount_new[i]*preDefGain[i])/(Ecount_new[i]) Where i represents the gray area in the color space, and count_new[i] indicates that the cumulative value of each gray area is a new cumulative value after the point of the overlap area is deleted according to the deduction value, PreDefGain [i] The gain value preset for each gray area. • 7. The automatic white balance correction method described in claim 1, wherein the color space displays the relationship of the ratios of red, green, and blue among the elements of the image. 8. The automatic white balance correction method as described in claim i, wherein the preset gain values of the plurality of gray regions are introduced for all light sources in the color space. °° 9· An automatic white balance correction method, comprising: inputting an image; introducing a color space and setting a plurality of gray regions in the color space to set a gain value; # analyzing the image, including extracting red of each element , green, blue value; calculate the falling point of each element in the color space in the image, and break the falling point of each gray area; introduce a point for deducting a point in an overlapping area between each gray area Deducting the ratio value; obtaining a point in the image where the pixel is in the overlapping area; deleting a part of the point falling on the weight region according to the deduction ratio value; 15 201032602 accumulating the deleted part after the point of falling a number of falling points in each gray area, and a plurality of integrated values corresponding to the plurality of gray areas are obtained; a new gain value is calculated by using the plurality of accumulated values and the plurality of preset gain values; and finally according to the new The gain value performs an automatic white balance procedure. 10. The automatic white balance correction method according to claim 9, wherein the new gain value is obtained by the following equation: (ScounUiewWPreDefGain^/^ount job (1)), where i represents a gray area in the color space , eGunt_new[i] indicates that the accumulated value of each gray area is deleted according to the deduction ratio value, and the accumulated value after the point of falling in the area is PreDefGain(1) is the incrementer of the P-page. & (2 £ U • The automatic white balance school as described in claim 9 wherein the color space displays the relationship of the ratios of red, green, and color in each element of the image. $ each 12. Please refer to the 9th item of the full-time _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The value of the number of gray areas,