TW200536376A - Method of noise reduction and image color interpolation - Google Patents

Abstract

A method of noise reduction is provided. A square pixel array is extracted from the pixel array on the sensitization device. Then a plurality of signal of pixels having the same color with the signal of a central pixel is extracted from there and the difference between the signals is judged that is too small or not. When the difference between the signals is not larger than a specific value, perform a sequential operation to get a preferred average of the signals to replace the signal of the central pixel, so as to blur the noise and the output image can gain a preferred quality.

Description

200536376 12223twf.doc 玖 、 Explanation of the invention: The technical field to which the invention belongs The present invention relates to a method for improving a picture image output by an image capturing device, and more particularly to a method for eliminating noise and an image color interpolation method . In the prior art, common image capture devices (such as digital cameras, etc.) usually use a charge coupled device (CCD) or a complementary metal oxide semiconductor (CMOS) as the light sensing device. However, the light sensor can only detect the intensity of the light and cannot determine the wavelength of the light. In order to take color photos, Tongdang will cover the photosensitive element with a layer of color filter. Color calenders only allow specific wavelengths of light to pass through. Therefore, as long as we read the signal of the photosensitive element (CCD or CMOS) located below the color filter, and add the known color of the color filter, we can Determine the wavelength and intensity of the light received by this pixel. However, since each pixel allows only a specific wavelength of light to pass through, only a certain wavelength of light can be read in a pixel, such as red, green, or blue. But in fact, each pixel in the picture must be composed of three colors of red, green and blue in order to match the actual color. Therefore, we must estimate the unknown color in a pixel. On the other hand, due to the inherent defect of the CCD or CM0S photosensitive element, there will be accompanying noise when the signal is generated, and the noise will affect the quality of the photos we take. Therefore, noise must be eliminated to improve the quality of the picture. 200536376 12223twf.doc It is known that a method of eliminating noise is to select a signal of the same color as the signal of the central pixel in a square pixel array, and average it to replace the original signal of the central pixel. However, another method of eliminating noise is to replace the original signal in the central pixel with the middle of these signals in the same color as the signal of the central pixel. It is also known that another method is to first delete the poles, that is, the minimum and maximum values in these signals of the same color as the signal of the center pixel, and then take the average of the remaining signals to replace the center. The original signal in the pixel is 値. However, these conventional noise elimination methods are still not enough to minimize the noise generated by the photosensitive element. SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a method for eliminating noise ', which can effectively reduce noise to improve the quality of the final image output by the image capturing device. Another object of the present invention is to provide an image color interpolation method, which can calculate each wavelength of light which must be in each pixel, so that the picture output by the image capturing device displays an approximate real color. The invention proposes a method for eliminating noise, which is suitable for eliminating noise generated by a photosensitive element, wherein the photosensitive element has an ax b pixel array. The method of eliminating noise is to first select an nx n pixel array arbitrarily from the ax b pixel array, and the center of this ηχ η pixel array has a central signal 値 Rx, y. Then, a plurality of signals 同 of the same color as the central signal 値 Rx, y are selected from the nx η pixel array to form a first set. Then the signals in the first set 値 200536376 12223twf.doc were divided into the first group and the second group multiple times. Then calculate the average 値 of the signal 値 in the first group and the second group respectively to obtain the first average 値 and the second average 値. After that, the absolute value of the subtraction between the first average value and the second average value is calculated, and the maximum value is used as the number of the first discrimination index. Then, it is judged whether the number of the first discrimination index is larger than the preset threshold 値, and when the number of the first discrimination index is larger than the preset threshold ’, the action of eliminating noise is ended. When the number of first discrimination indicators is less than the preset threshold 値, the average signal 値 is calculated. Finally, this signal average 値 is used to replace the central signal 讯 Rx, y. Among them, the so-called signal average 値 here refers to the average of some signals 値 in the ηχ η pixel array. According to the embodiment of the present invention, the method for differentiating the signal signals in the first set into the first group and the second group multiple times includes first arranging the signal signals in the first set from small to large to form The first ordered set then makes the signals of items 1 to i_l in the first ordered set 値 the first group 6 (/), and the items i to k in the first ordered set The signal 値 is the second group 02 (0. Among them, k is the total number of signal 値 in the first ordered set, and i is an integer, and 2 ′ i $ k-2. According to an embodiment of the present invention As mentioned, the first average 値 is α1 (0, the second average (is% (/), and% (〇 | = / (〇 ', and the number of discriminant indicators is the largest 値 of j (/). According to the present invention According to the embodiment, the method for calculating the average signal 値 includes firstly calculating and obtaining a plurality of differences ， dx ,, y, where Rx ,, y, represents the signals 値 in the first set. Then these differences値 dx ,, y, form the second set, and then divide these differences 値 dx ,, y in the second set into the third group Q (0 and the fourth group 6 (0. meter Calculate the average of the third group g3 (/) by 200536376 12223twf.doc and the fourth group G4 (these differences 値 dx ,, y, within 〇, and get the third average 値 < 0 and the fourth average 値 α4 (〇. Then calculate the number of the second discriminative index / (>, where ⑺ = give:} guide, and Go) and Lat) are the third group like / ·) and the fourth group G4 (/ · ). Finally, the second discriminative index number P (0 is the largest 値, and the average of the signal 値 corresponding to the third group 6 (0 within the range 〇) is calculated at this time, and According to the embodiment of the present invention, the method of dividing the difference in the second set into the third group 〇3 (〇 and the fourth group G4 (〇) includes firstly dividing the second set The median rates are arranged from small to large to form a second ordered set, and then the rates of items 1 to i-1 in the second ordered set are the third group g3 (0, and the The signals 第 to 第 in the two ordered sets are the fourth group G4 (0. where k is the total number of differences in the second ordered set, and i is an integer, and 2Si $ k-2. According to the embodiment of the present invention, The method for removing noise further includes calculating the number of third discrimination indicators from the signal 値 in the first set, and then determining whether the number of third discrimination indicators is greater than the preset threshold 値. When the number of third discrimination indicators is greater than the preset threshold 値, then Stop the action of eliminating noise. When the number of the third discrimination index is not greater than the preset threshold 値, calculate the total average of the signal 値 in the first set to obtain the signal 値. Finally, replace the central signal with the signal 値値. According to the embodiment of the present invention, the method for calculating the third discriminant index number includes firstly arranging the signals 値 in the first set from small to large to form an ordered set of 200536376 12223twf.doc. Then, the average 値 of the first two signals 値 in the first ordered set is subtracted from the average 之后 of the last two signals in the first ordered set to obtain the ^ th discrimination index number. The invention provides an image color interpolation method, which is suitable for a photosensitive element having a Bayer pixel array with ax b. In this image color interpolation method, an ηχ η pixel array is arbitrarily selected from a Bayer pixel array of a X b, and a signal 値 is provided in a central pixel located at the center of the ^ X η pixel array. Then calculate the signal 値 change rates above, below, left, right, upper left, lower left, upper right, and lower right of the center pixel. There are at least one red light signal 値, at least one green light signal 値 and at least one blue light signal 値 in each direction. Then the formula T = k ^ x Min + k2x (Max-Min) is used to calculate the predetermined threshold 値 T. Among them, Min is the minimum of the change rate of these signals, Max is the maximum of the change rate of these signals, and h and k2 are real numbers. Then select some directions to form a set. Among them, the rate of change of the signals in these directions is smaller than the predetermined threshold. Then add the red light signal 値, the green light signal 値 and the blue light signal 集合 in the collection respectively to obtain the red light signal 値 total Gsum, the green light signal 値 total RsUm, and the blue light signal 値 total Bsum. Finally, the unknown green light signal in the central pixel is calculated from the red light signal 値 total Gsum, the green light signal 値 total, the blue light signal 値 total B_, the red light signals 値, the green light signals 値, and the blue light signals 値. value. According to the embodiment of the present invention, when the signal 値 of the central pixel is a red light signal 値 r5, the equation §5 = ι · 5 + ΟΙΙ · = ι · 5 + (0_-Κ3 _) / ^ ί3 can be obtained. Output g5, where g5 is the green light signal of the central pixel, and Sets is the number of directions in the set. 200536376 12223twf.doc According to the embodiment of the present invention, when the signal 中央 of the central pixel is a blue light signal 値 b5, the green light of the central pixel can be obtained by using the equation g5 two b5 + GBdiff = b5 + (Gsum-Bsum) / Sets Signal 値 g5. According to the embodiment of the present invention, when the signal 中央 of the central pixel is a green light signal, and the two blue light signals 値 and see ^ and the two red light signals 値 圪 7 and 绿 are respectively above the green light signal 値 & The signal in the bottom, left, and right pixels 値, the green light signal 値 is the signal in the upper pixel of the blue light signal 値, and the green light signal 値 is the blue light signal 値 See the signal in the pixel below the 値, green The light signal 値 G_2, y is the signal in the left pixel of the red light signal, especially ^, the green light signal 値 is the signal in the right pixel of the red light signal 値 Aw, and the green light in the ηχ η pixel array When the signal 値 is known, this image color interpolation method further includes calculating the difference 値 KR ′ of the red light signal 値 and the green light signal 中 in each pixel having the red light signal 其中 where (X -1, less) = Wide, 'and (X +1, less) = _ &+1;; ° Then calculate the weight ice of each difference 値 KR in its corresponding pixel, where

~ 1, less = / 1 + V 2 | (GX, y " Gx ^ 2) ~ 2 〇1 2 J ten 1 2 J / then calculate the blue light signal 値 and the green in each pixel with these blue light signals 値The difference between the optical signals 値 KB, where =, and 仏 (x,} + l) = C ^ + 1-. Then ’calculate the weight w of each rate KB in its corresponding pixel, where 200536376 12223twf.doc

Then use the formula ΚΏ (χ, ν) =: wx-ly + KR (x + and KB (x, y) = £ i ^ lI) x ^ -i + Kb (x ^ + 〇x ~ -1 + %, "1 to find the difference between the central pixel and the difference 値 A (χ, β. Then use the difference between the central pixel (χ, >)) and the central pixel's green light signal 値 (^ 〃 to infer The central pixel's red light day is only 5 tigers. On the other hand, the difference between the central pixel and the central pixel's green signal 値 G〃 is used to infer the central pixel's blue signal 値 I, where 1 = ϋ (υ) According to the embodiment of the present invention, when the signal 値 of the central pixel is a blue light signal 且, and the red light signal 値 is a signal 値 in the upper left pixel of the blue light signal 値, the red light signal 値 7 ^ ㈣ is a blue light signal The signal 値 in the lower left pixel 値, the red light signal 値 & heart M is the blue light signal (see the signal in the upper right pixel of π), and the red light signal 値 is the signal in the lower right pixel of the blue light signal 値 ' The signal 値 in the upper left pixel of the red light signal 値 is a green light signal 値 G-2, y_2, the signal in the red light signal 値 7 ^, and the signal in the lower left pixel ^ is a green light signal 値Red signal signals the upper right of the pixel Zhi Zhi Zhi 200536376 12223twf.doc for the green signal

Gt + 2 ,,, _ 2, the red light signal 値 < + l v + 1 The signal in the lower right pixel 値 is a green light signal 且, and these green light signals n in the nX n pixel array are all known At this time, the image color interpolation method further includes first calculating the difference KR between the green light signal 値 and the red light signal 中 in the upper left pixel, lower left pixel, upper right pixel, and lower right pixel of the central pixel. Among them, 1, 1, less-1): = (^ 丨 广 1- & points 丨, hy-i 'ΚΝ (χ-Ιγ + ή = σχ_1γ + ι- ^ x-\, y + \ κκ (χ + Ιγ + \) = βχ + 1γ +]-^ x + \, y + \ Then calculate the weight W of each difference 値 KR, where ^ -1, less -1 vx-\, y + \ jc + l, less +1 1 + 1 + 1 + 1 +

G x + l, y + l ~ Gx one 1, less _1 -2, less -2 2

Gx + hy-l ~ Gx-ly + \ y _l_ (Gx, y-Gx-2, y + 2 ^ 2Λ Ten J 1 2 JJ Gx, y ~ Gx G score 丨 -1 丨, less + 丨 x, y ^ x + 2, y + 2 and then the equation

Wx ~ \, y ~ \ + ^ + ^ + 1, ^ + 1 12 200536376 12223twf.doc Find the difference between the central pixels. Finally, the difference between the central pixel 値 and the central pixel's green signal 値 G〃 is used to infer the central pixel's red signal 値; ^, which is calculated as A (xj). According to the embodiment of the present invention, when the signal 値 of the central pixel is a red light signal 値 and the blue light signal 値 see-㈤ is the signal 中 in the upper left pixel of the red light signal 値 7 ^, the blue light signal 値 5 ^, — Is the signal in the lower left pixel of the red light signal, and the signal heart in the blue light signal -i is the signal in the upper right pixel of the red light signal 7 ^, see the +1 for the blue light signal, and y + 1 is red The signal in the lower right pixel of the light signal 値 7 ^, and the signal in the upper left pixel of the blue light signal 値 See ^ is the green light signal 値 G_2, y_2, and the blue light signal is in the lower left pixel ^ + 1. The signal 値 is the green light signal 値 G_2 ,, + 2, the blue light signal 上方 is the signal in the upper right pixel 値 is the green light signal 値 G + 2, 2, _2, the blue light signal mode + 1 > +1 the lower right pixel When the signal 値 in the image is a green light signal ^ x + 2, y + 2 and the green light signals 値 in the ηχ η pixel array are all known, this image color interpolation method further includes first calculating the upper left of the central pixel Difference between green light signal and blue light signal in pixels, bottom left pixels, top right pixels, and bottom right pixelsWhere KB (x-hy-l) = Gx KB {x ^ \, y-\) = Gx + Xy _]-Bx + X y_x ^ especially "you-1, less + 1) = (^-丨, state -足 七 + 丨 'Α (χ + 1,3; + 1) = (^ + 1 ”丨 Again 丨 〆13 200536376 12223twf.doc Then calculate the weight w of each rate 値 KR, where

-1, less -I f 1 + 1, ^ + 1 ~ ^ .ν-Ι, ν-Ι 2 丄 (Gx, y '-GU, Guang 2, 2, 1 but V 1 2 J ten \ 2 J)

Jx-ly + l x + \, y ~ \

Gx, y '~ Gx \ 2 \ +2, /-, 2

\ +1, state, 2 1 +

(1 -L (r r Λ ^ χ + \, γ- \ ~ Kjrx-\, y + \ 2 丄 (GXJ-Gx_ly + 1 \ 2 \ 1 ding V {2 J ten l 2 J J 1 +

(a -G WJC-1, one less 丨 WJC + 1, less +1 ^ GXiy ^ Gx ^ y ^ 2 and then use the equation

Kb (upper x)-&x; l less one + Α (χ + 1, less one l) x> ^ + ljM + Wx + lyA + &^; ^ + ι, ηι Find the difference between the central pixels 値 Α (X, β. Finally, the difference between the central pixel and the green light signal 値 Gy of the central pixel is used to infer the blue light signal 値 L of the central pixel, and the calculation method thereof is L = G ^ -Uw). The noise is eliminated, and then the color interpolation is performed, so that the noise can be prevented from spreading during the color interpolation, thereby improving the quality of the picture. In order to make the above and other objects, features, and advantages of the present invention more obvious and easier Understand, a preferred embodiment is given below, and will be described in detail in conjunction with the attached drawings. 14 200536376 12223twf.doc Implementation The present invention uses special methods to eliminate noise and color interpolation, respectively, so that The image output by the image capture device has a high degree of similarity without distortion to the captured image. The following embodiments are described by taking a 5 × 5 pixel array as an example, but those skilled in this art should know that this The method disclosed in the invention can also be applied to other pixel arrays. The invention The size of the pixel array is not limited. Figure 1 does not show a 5x5 pixel array 100 arranged in a Bayer pattern. Please refer to the figure first, where R, G, and B represent the reading of the photosensitive element (not shown). The red light signal 値, green light signal 値, and blue light signal 値 obtained, and the subscripts R, G, and B indicate the position of the pixel where they are located, which are centered on the signal 値 Rxy of the central pixel. The subscripts are increased or decreased to the left and right (X axis) and up and down (y axis). The pixel array 100 shown in FIG. 1 is taken as an example to illustrate the flow of the method for eliminating noise in the present invention. 2 is a flowchart of steps of a method for eliminating noise according to a preferred embodiment of the present invention. Please refer to FIG. 1 and FIG. 2 at the same time. In step S200, an ax b pixel array (not shown) on a photosensitive element is firstly The pixel array 100 is selected. Among them, a may be equal to or not equal to b. In other words, the pixel array on the photosensitive element may not be a square array or a square array. Then, as described in step S202, the pixel array is Select from 100 The central signals 値 Rx, y are signals of the same color, which are, for example, red light signals. Then step S204 is performed to divide the selected red light signals 値 15 200536376 12223twf.doc into a first group and The second group. In more detail, in step S202, the red light signal 値 is first selected and -2, ^ -2, 乂 1, and L㈣ are selected, and then these red light signals 値 are formed into the first set S Among them, in S204, the red light signals 値 are first sorted from small to large to form the first ordered set rank (S). Wherein brain A: {= {7 ?. , 7? 丨, $ 2, /? 3, and 4, ^? 5, ^? 6, and 7}. It can be seen from this that the size of the red light signal 値 in the first ordered set rank (S) is Fan < Ruler < 7? 2 &7; 3 < Fan < Exact < 7 & 6 < Exhaust. Then, select the red light signals from the first to the first order set ^ 1 ^ (8) to red light signals 値 to the first group Gl (〇, and the red light signals from the i to k items値 is the second group q⑺, where 2 S i Sk-2, and k is the total number of red light signals 中 in the first ordered set rank (S). In this embodiment, k is equal to 8 for example. It should be noted that the red light signal 値 R in the first ordered set rank (s) is used to distinguish each red light signal 値, and the red light is sequentially expressed by an integer from 0 to 7. The size of the signal 値 is not related to the subscripts previously indicated by x and 7, respectively. Thereafter, as described in step S206, the average of the first group ^) and the second group G2 (/) is calculated to obtain The first average number “1⑺” and the first average number “2⑺” are obtained. Among them, the calculation formula of the first average number% (/) is force (0 = @ ', the second average number α2 (〇 is ten ten ti). Is a2 (0 = y7 ^ A ° K-1 j = i 16 200536376 12223twf.doc Then proceed to step S208 to find the first discriminant index number, which is outside the equation first) = h⑺-α2 (/) | one by one Find J (2), J (3) ... J (6), and its The largest threshold is the number of the first discriminant index. Then judge whether the number of discriminant indicators * is greater than one ^ preset threshold, as described in step S210. The preset threshold here is after multiple experiments. The obtained experience 値. In step S210, when the number of the first determination index is greater than a preset threshold 値, the work of eliminating noise is stopped. When the number of the first determination index is less than a preset threshold 値, the process continues. In step S212, the average signal 値 is calculated. Finally, in step S214, the central signal 値 Rx, y is replaced with the average signal 値, and the noise is eliminated. In step S212, the algorithm of the average signal 系 is calculated first. ， / = | 尤, yl | to find the majority of the difference 値 dx ,, y, 'where Rx ,, y, represents each red light signal 値 in the first set S. Then, the difference 値 dx ,, y, Make up the second set D, where /) = ½ -2, less -2, 々 > / -2, 4 + 2, less -2, "+2, less, 4-2, less +2, 々y +2, 4 + 2, less +2} 〇 Then the difference in the second set D is divided into a third group G3 (0 and a fourth group G4 (0. The method is as described above, The difference haiku ^ in the second set D is arranged in order from small to large to form a second ordered set rank (D), where moxibustion (1)) = {^ /., ^ / 丨, ^ / 2, (^ 3, (^, 4, ^ / 6, (^ 7). Then select the second ordered set ^ 111 < :( 0) from the 1st to the i-Ith difference rate is the first A group, and the difference between the i-th and k-th terms is the second group G2 (0. Of course, i and k here are the same as described above. Then calculate the average of the difference between the third group q (0 and the fourth group g4 (0), and obtain the third average 値 α3 (0 and the fourth average 値 α4 (0. The formula is 17 200536376 12223twf. doc Gu) =} §meaning, 々0 = ^ 1 ^, where ⑽ represents the difference in the second ordered set rank (D). Then use the formula to find the increase in the number of the second tide index. And from) are the variation numbers of the difference between the second group G3 (0 and the fourth group, respectively, the formula is Gamma = α3 | 2, and. • = 0 and then calculate ρ respectively (2), P (3) ... P⑹, to obtain the maximum 値 of the second discriminative index number ⑺, and calculate the average of the signal 値 corresponding to the third group (? 3 (0 within the difference dj, The average signal 値 is obtained. For example, if P (5) is the maximum 値 of the second discriminant index number P (0, then the red light signals 値 corresponding to heart, heart, d2, d3, and d4 are obtained, respectively. Then, calculate the signal average 値 of these red light signals 最后 Finally, as described in step S214, replace the central signal 値 Rx, y with the signal average 算出 calculated in step S212, thus ending the process of eliminating noise in this embodiment. In addition, After inventing the above-mentioned method for noise reduction to reduce noise in the photosensitive element, another noise reduction process can be performed 'in order to further improve the quality of the output picture. Another embodiment will be described below at this stage 3 is a flowchart of a method for eliminating noise in another preferred embodiment of the present invention. Please refer to FIG. 1 and FIG. 3 at the same time. The pixel array 100 is selected from the ax b pixel array. Then, as described in step S302, a signal 同 having the same color as the central signal 値 Rx, y is selected in the pixel array 100, for example, a red light signal 値 18 200536376 12223twf.doc

Rx-2, y_2, Rx, y-2, K + 2, y-2, hy, Rx + 2 y, Kx_2 y + 2, y + 2, and Rx ... + 2, then make these red light signals 値Compose the first set S, where. {义 -2 ^ -2, & "person 2, with 4-2, / + 2, Yichi +2,7? ^ + 2,7? ^ + 2}. Then in step S304, follow this first The size of these red light signals 値 is sorted from small to large to form the first ordered set rank (S). Among them, ra ^ :( 5 *) =. It can be seen that the ordered set rank ( The size of the red light signal in S) is ~ < R '< R2 < R3 < R4 < R5 < R6 < Rj. Then, in step S306,' select the ordered set ranj ^ S). The two largest signals 値 R6, R? And the smallest two signals 値, R! Are calculated and subtracted respectively to obtain the third discriminant index number 2 (/). The formula is 2⑺ = (( Where + exhaust) / 2- (heart + catty) / 2). In step S308, it is determined whether the number of the third determination index ρ 大于 is greater than a preset threshold 値. When the number of the third determination index ⑽ is greater than the preset threshold, Then the action of eliminating noise is stopped. However, when the number of the third discrimination index ρ (0 is not greater than the preset threshold 値 Τ), the process proceeds to step S310, and all signals in the first set S are added to average, and thereafter Then replace the original central signal 値 Rx with the obtained average signal 所得, y, as described in step S312. At this time, the process of eliminating noise in this embodiment is completed. After the noise generated by the photosensitive element is eliminated, the color interpolation of the image can be continued. The color interpolation can be divided into There are two cases, one is to interpolate unknown green, and the other is to interpolate unknown red or blue. In the present invention, the color interpolation is also divided into two stages. In the first stage, all unknowns are calculated first. Green, the remaining unknown red and blue are calculated in the second stage. The detailed method will be described in the following example 19 200536376 12223twf.doc. Figure 4 shows a 5x5 pixel array 400 arranged in a Bayer pattern. The known color of the signal 値 in each pixel is shown in Figure 4, where r is the red light signal 値, g is the green light signal 値, and b is the blue light signal 値. The numbers marked in Figure 4 are for Differentiate the signal 値 in each pixel. Please refer to FIG. 4. The signal 値 in the central pixel of the pixel array 400 is r5, which shows that the red light signal 値 is known here, so first perform a calculation to find the central pixel. Green light No. Zhi. First, r5 where the center pixel, which is calculated on the rate of change of the signal eight directions, which follows the formula

Gradient N = | g4-g9 | + | r2-r5 | + | bl-b3 | / 2 + | b2-b4 | / 2 + | gl-g6 | / 2 + | g2-g7 | / 2

Gradient E = | g7-g6 | + | r6-r5 | + | b2-bl | / 2 + | b4-b3 | / 2 + | g5-g4 | / 2 + | g10-g9 | / 2 Gradient S = | g9-g4 | + | r8-r5 | + | b3-bl | / 2 + | b4-b2 | / 2 + | gl l-g6 | / 2 + | gl2-g7 | / 2

Gradient W = | g6-g7 | + | r4-r5 | + | bl-b2 | / 2 + | b3-b4 | / 2 + | g3-g4 | / 2 + | g8-g9 | / 2

Gradient NE = | b2-b3 | + | r3-r5 | + | g4-g6 | / 2 + | g7-g9 | / 2 + | g2-g4 | / 2 + | g5-g7 | / 2

Gradient SE = | b4-bl | + | r9-r5 | + | g7-g4 | / 2 + | g9-g6 | / 2 + | gl0-g7 | / 2 + | gl2-g9 | / 2 Gradient NW = | bl-b4 | + | rl-r5 | + | g4-g7 | / 2 + | g6-g9 | / 2 + | gl-g4 | / 2 + | g3-g6 | / 2 Gradient SW = | b3-b2 | + | r7-r5 | + | g6-g4 | / 2 + | g9-g7 | / 2 + | g8-g6 | / 2 + | gl l-g9 | / 2 where N, S, W and E represent The upper, lower, left, and right sides of r5, and NE, NW, SE, and SW] respectively represent right_h, upper left, lower right, and lower left of r5. Then use the formula T = M Min + k2x (Max-Min) to calculate the predetermined threshold 値 T, where Min is the minimum of the 8 signals 値 change rate, Max is the maximum of the 8 signals 値 change rate, and 1 ^ And 1 ^ 2 is a real number. And in the present embodiment, 4 cases 20 200536376 12223twf.doc, such as Daiji Temple 1.5 · 1 <: 2 are equal to 0.5, for example. Then, among the above eight signal change rates, a direction ′ whose 値 is smaller than the threshold 値 τ is selected to form a set Η. For the convenience of explanation, we assume that one of the signal change rates in these 8 directions is shown in Table i: Direction Up Right Bottom Left Right Top Right Bottom Left Top Left Bottom Signal Change Rate 12 13 7 8 4 7 12 14 Table 1

Therefore, the threshold 値 is τ = 1 · 5χ 4 + 0 · 5χ (14-4) = 11, and then the direction in which the signal change rate 値 in eight directions is less than n is selected to form the set η. Therefore H = {S, W, NE, SE}. Then we build the signal in each direction in the set 値 according to the color to build Table 2:

& light: signal 値 blue light signal 値 red light signal 値 g9 (b3 + b4) / 2 (r5 + r8) / 2 left g6 (bl + b3) / 2 (r4 + r5) / 2 top right (g2 + g4 + g5 + g7) / 4 b2 (r2 + r3 + r5 + r6) / 4 top left (g7 + g9 + gl〇 + gl2) / 4 b4 (r5 + r6 + r8 + r9) / 4 Table 2 From Table 2 It can be seen that there are green light signals 値 g9, blue light signals 値 b3 and b4 I, and red light signals 値 r5 and r8 in the direction below r5. Then we add the signal signals 相同 of the same color in 21 200536376 12223twf.doc to find the sum Gsum, Bsum, and Rsum of the green light signal 値, blue light signal 値, and red light signal 分别, the formula is as follows:

Gsum = g9 + g6 + (g2 + g4 + g5 + g7) / 4 + (g7 + g9 + g 1 0 + g 12) / 4 ®sum ~ (b3 + b4) / 2 + (bl + b3) / 2 + b2 + b4 Rs_ = (r5 + r8) / 2 + (r4 + r5) / 2 + (r2 + r3 + r5 + r6) / 4 + (r5 + r6 + r8 + r9) / 4 Then use the known The red light signal 値 r5 and the calculated green light signal 値 sum Gsum and the red light signal 値 sum Rsum are used to find the g5 of g5. The formula is: g5 = r5 + GRdiff = r5 + (Gsum-RsUm) / Sets where, Sets 値 represents the number of directions in the set H. In this embodiment, Sets is equal to 4, for example. The obtained g5 is the green light signal 値 in the central pixel. In addition, if the signal 値 known by the central pixel is the blue light signal 値 b5, then the above formula for obtaining g5 is changed to: g5 = b5 + GBdiff = r5 + (Gsum-Bsum) / Sets and the calculation flow before this formula is It is the same as the above description, and will not be repeated here. After the green light signal 値 of each pixel in the pixel array 400 is obtained by the above algorithm, interpolation of the unknown blue light signal 红 or red light signal 値 in all pixels can be continued. In the process of interpolating the blue light signal or the red light signal, it can be divided into two examples to illustrate. The examples will be used to explain these two examples. Example 1 FIG. 5 shows a 5 × 5 pixel array 500 with a partial Bayer pattern arrangement. In the pixel array 500, the known signal 値 in the center pixel is a green light signal 値 Gx, y, and among the four pixels adjacent to the green light signal 値 Gx, y, 22 200536376 12223twf.doc is a known signal値 is see, M, Jiu, ^, You Qi and You +1, respectively, and the signals 値, 1. + 1, and ruler 贝贝 [J and green light signals 値 ^ -2, ^ + 2, respectively And (? _ Adjacent, as shown in Figure 5. Please refer to Figure 5, if you want to interpolate the unknown red light signal 値 of the central pixel, the red light signal 能够 that can be referenced is only the left and right sides. However, since In the algorithm of the first stage of the invention, the green light signal 値 in all pixels has been calculated. So at this stage, the green light signal 所有 of all pixels is known. Therefore we can use the known red light The signal 値, blue light 値, and green light 値 obtained in the first stage are used to calculate the & sum of the four pixels adjacent to the central pixel. The calculation method is as follows: KR {X- ^ y) = Gx- hy-Rx-ly KR (x + 1? = Gx + ly-Rx + hy especially β (X, less-1) = Gjcj-I-see, _V-1 A (especially, less + 1) = Gjc, y +1 ~ Ί_ν + 1 This recalculation down four pixels of KR * Kb weights w, the formula is

W x-ly ^ + 1, less meaning, less one 1 1 + r GXJr, -Gx x + \, y -yUV fly ^ l ^ Gx-\, y ~ Gx ^ 1 + + (Gx, y ^ Gx-2 , y (Gx, y ^ Gx + 2, y G ^ y ^ Gx, y-2

Gx, y ^ \ ~ Gx, y- \ 2 A less +1

f 1 + l 2 JV • f G ^ " ~ G ^ y 23 200536376 12223twf.doc We then multiply Kr and Kb by the corresponding weights to find the sum of Kb at the central pixel: ^ -v-Ι, ν + KB (x, y) = 〇x 〇x '跔 丨 + ~ + 1 Finally, use the ruler 11 and ruler 1) above and the green light signal 値 Gy whose central pixel is known to calculate the center. The red light signal 値 ^^ and the blue light signal 値 in the pixel are shown in the figure below. The calculation method is as follows:

Rx, y = Gx, y-KR (x, y)

Bx, y = G ^ y-KB (^ y) Example 2 FIG. 6 shows a 5 × 5 pixel array 600 with a partial Bayer pattern arrangement. In the pixel array 600, the known signal 中央 in the central pixel is a blue light signal 値 Bx y, and among the four diagonal pixels in a 3x 3 pixel array centered on the blue light signal 値 Bx y, the known signal 値Red light signals 値 An, & L ..., and L㈣, while green light signals 値, G.a + 2, and A + 2, and γ2 are respectively located in a 5x5 pixel array centered on the blue light signals 値 Bx, y. Of the four diagonal pixels, as shown in Figure 6. Please refer to FIG. 6. Since the green light signals 所有 in all the pixels of the pixel array 600 are obtained in the first stage calculation process, the green light signals 値 Gx, y in the central pixel of the pixel array 600 are also It is known, and only the red light signal Rx, y is unknown in the central pixel. Therefore, in this stage, ie 24 200536376 12223twf.doc, the KR can be calculated using the green light signals 値 and 圪 heart-, and the meaning + ^ obtained in the first stage. The calculation method is as follows: KR-1,-l) = Gx-\, y- \-Rx-\, y- \ ~ (x + l, less -1) = ^ x + \, y- \ ~ ^ x + \ , y- \ kr (^ ~ Ly + l) = Gx_ly +]-Rx. \, y + \ KR (x-hl, y + l) = Gx + l y +]-^ x + \, y +] The weight of the four heart palpitations W, its formula is 1, 2, 2, and 1 + x, y x-2, y-2 2 ^ x- > r \, y + \ ^ x-\, y- \ 2 vx + \ yy- \

(1 + V 2 4- ^ x, y ~ ^ x + 27y ~ 2 2, l 2 J 丁 l 2 JJ vx-\, y + \ f 1 + V 2 -L (GX ^ -GX ^ A 2, l 2 J 丁 l 2 J / yx + \, y + \ f 1 + V (Gx-ly- \ -Gx + ly ^ 2 4- rGx, y-Gx + 2iy + 2 ^ 2, {2 JI {2 J y

Then multiply the & of these four pixels by the corresponding weights to get the central pixel's A (, ^ (x --l) x + + [A less '_ I… + +% 々 + 1 +' +1 , KR (x -1? .Y + 0X + KR +1,7 + l) x Wx +] y + l% _ 丨, Guang 丨 + ~ Guang 丨 + > v 丨, and finally use the calculated & 値 and the known green light signal 値 Gx, y ', the red light signal 値 RX, y in the central pixel 25 200536376 12223twf.doc can be found, and the calculation method is as follows: In addition, if the signal of the known central pixel 値 is The blue light signal 値 Bxy, as shown in the pixel array 7000 shown in FIG. 7, first calculates the ^ 1 ^ using the green light signal 値 and, 1 ^ 1, and 5_ + 1 obtained in the first stage. As follows =-0 = Gx-ly- \-Bx-ly-l

KB {x ^ ly- \)-Gjf + 1, less -1-Ar + 1, less -1 KB (x-\, y + \) = Gx_ly +] -Bx_ly +] You Zhao (X + 1, less + 1 )-^ jc + l, y + l ^ x + \ yy + \ Next calculate the weights w of these four KB 値, the formula is: 4-1, less -1 1 +

G x + \, y + \

Gx -Uy- 2, less -2 ν '% + 1, wide 1 = 1 + frr \ 2 4- (Gx, y 2Λ l 2 J \ \ 2 J y KJC-l, y + l (1 + \ r Gx ^ v- \-~ Gx-\, y ^ 2 4. GU, "2, 2 \ ~ 2 {2 J 卞 l 2)) —Gx

Then multiply the KB of the four pixels by the corresponding weight to get the KB of the central pixel

vx +], y + \ 1 +

G i-l, less -1 +1, less +1

V

(GY -G meaning, less x + 2, y + 2 26 200536376 12223twf.doc h (x ~ \, y ~ \) x wx_x + KH {x + \, y ^ \) x wx, x ^ x vx- \, y- vx + \, y +] [KB (x-\, y + \) xwx_Xy ^ ^ KB (x + \, y + \) xwx + x ^ x

Finally, using the calculated 仏 値 and the known green light signal 値 Gx, y, the red light signal 値 Bx, y in the central pixel can be obtained. The calculation method is as follows: Βχ, γ = 0χ ^ -ΚΒ (χ ^ ) In summary, the present invention has the following advantages: _ 1. Effectively reduce the noise generated by the photosensitive element, so that the output picture has better quality. 2. Accurately calculate the signal 値 of each color in each pixel to improve the similarity between the color of the output picture and the real object, and thus avoid distortion of the picture image. 3. The noise elimination process of the present invention can eliminate noise before performing image color interpolation, so that the noise can be prevented from spreading during the color interpolation process, thereby improving the quality of the tube image. 0 Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications and retouching without departing from the spirit and scope of the present invention. The scope of protection of the invention shall be determined by the scope of the attached patent application. [Brief Description of the Drawings] FIG. 1 shows a 5x5 pixel array 100 with an arrangement of ~ Bayer patterns. 27 200536376 12223twf.doc FIG. 2 shows a flowchart of a method for eliminating noise in a preferred embodiment of the present invention. FIG. 3 is a flowchart illustrating steps of a noise elimination method according to another preferred embodiment of the present invention. FIG. 4 illustrates a 5 × 5 pixel array 400 arranged in a Bayer pattern. Figure 5 shows a 5x5 pixel array 500 arranged in a partial Bayer pattern. Figure 6 shows a 5x 5 pixel array 600 arranged in a partial Bayer pattern. Figure 7 shows a 5x 5 pixel array 700 arranged in a partial Bayer pattern. . [Illustration of Graphical Symbols] 100, 400: 5x 5 pixel array S200 arranged by Bayer pattern: 5x 5 pixel array S202 is selected: Signal of 5x 5 pixel array with the same color as the center signal is selected S204: These are selected The red light signal is divided into the first group and the second group S206: calculating the first average number) and the second average number seven. S208: Finding the number of the first discrimination index S210: Whether the number of the first discrimination index is greater than a predetermined number Set the threshold 値 S212: Calculate the signal average 値 S214: Replace the central signal with the signal average 値 S300: Select a 5x5 Bayer pixel array 28 200536376 12223twf.doc S302: Select a 5 × 5 Bayer pixel array with the same I! Color as its central signal Signal 値 S304 · Sort these signals from small to large according to the size of these signals 形成 to form a first ordered set S306: Find the number of the third discrimination index S308: Determine whether the number of the third discrimination index is greater than the preset gate number S3 10: All signals in the first set are added and averaged to obtain the signal average. S312: Signal average is used instead of the central signal. 500, 600, 700: 5 × 5 pixel array with partial Bayer pattern arrangement

29

Claims (1)

200536376 12223twf.doc The scope of patent application: 1. A method for eliminating noise, suitable for eliminating noise generated by a photosensitive element, wherein the photosensitive element has an ax b pixel array, and the method for eliminating noise includes: An ηχ η pixel array is arbitrarily selected from the ax b pixel array, and a central signal 値 Rx y is provided in the center of the ηχ η pixel array; a plurality of signals 同 having the same color as the central signal 値 Rx, y are selected to form a The first set; dividing the signals 将该 in the first set into a first group and a second group multiple times; calculating each time in the first group and the second group The average 値 of the signals 値 is obtained as a first average 値 and a second average 値; each time the absolute 値 of the first average 値 and the second average 减 is calculated, and the maximum 値 is taken as A first discriminative index number; judging whether the number of the first discriminant index is greater than a preset threshold 値, when the number of the first discriminant index is greater than the preset threshold 値, stop the action of eliminating noise; index When the number is smaller than the preset threshold 値, a signal average 値 is calculated, where the signal average 値 is an average of a part of the signal 値 in the ηχ η pixel array; and the signal average 値 is used to replace the central signal 値 Rx, y. 2. The method for eliminating noise according to item 1 of the scope of the patent application, wherein the method of dividing the signals in the first set into the first group and the second group multiple times includes: 30 200536376 12223twf.doc arranges the signals 値 in the first set from small to large to form a first ordered set; and makes items 1 to i-1 of the first ordered set Some of the signals 値 are the first group G, (0, and some of the signals i to k in the first ordered set are the second group 02 (0, where k is the total number of the signals 値 in the first ordered set, and i is an integer and 2 $ i $ k-2. 3. The method for eliminating noise as described in item 2 of the scope of patent application, Wherein the first average 値 is < /), the second average 値 is α2 (〇, and ® h (/)-β2 (〇 卜 J (〇, the number of the first discriminant index is the largest of J (〇) 4. The method for eliminating noise as described in item 1 of the scope of patent application, wherein the ηχ η pixel array includes a 5x5 pixel array. 5. The elimination as described in item 1 of the scope of patent application The method of calculating the signal, wherein the method of calculating the average signal V includes: Calculating You V = | You, and many more rates dx ,, y, where Rx ,, y, represent the signals in the first set 値; Divide the rates 値 dx ,, y into a second set; _ divide the rates 値 dx ,, y in the second set into a third group 6 (0 and a fourth Group g4 (0, where i is used to distinguish the third group and the fourth group each time; calculate the third group G3 (0 and the fourth group 64 (0 within The average of these differences 値 dx ,, y, to obtain a third average 値 α3 (/ ·) and a fourth average 値 α4 (ζ ·); calculate a second discriminant index number π), where the voice and (0 + ^ 2 (0 31 200536376 12223twf.doc A (/) and ¥ 0 are the variation numbers of the rates within the third group G3 (0 and the fourth group g4 (0); and Calculate the maximum value of the second discriminative index number eight /), and calculate the average of the signal values corresponding to the rates in the third group G3 (at this time, the signal average value) 6. If applying for a patent The method for eliminating noise described in item 5, wherein the rates dx ,, y in the second set are distinguished into the third group G3 (/) and the fourth group G4 multiple times. The method of (0) includes: arranging the rates dx ,, y in the second set from small to large to form a second ordered set; and making the first item in the second ordered set The rates from part i to I-1 are the third group G3 (0, and parts from i to k in the second ordered set are the fourth group. Group G4 (0, where k is the total number of rates in the second ordered set, and i is an integer, and 2 $ iSk-2. 7. The method for eliminating noise as described in item 1 of the scope of patent application, further comprising: calculating a third number of discriminant indicators from the signals in the first set; judging whether the number of the third discriminant indicators is greater than A preset threshold 値, when the number of the third discrimination index is greater than the preset threshold 値, the action of eliminating noise is stopped; when the number of the third discrimination index is less than the preset threshold ，, the number in the first set is calculated The signal average is obtained by the total average of the signal signals; and 32 200536376 12223twf.doc replaces the central signal signal with the signal average signal. 8. The method for eliminating false information as described in item 7 of the scope of patent application, wherein the method for calculating the number of the third discriminant index comprises: arranging the signals in the first set from small to large. A first ordered set; and subtracting the average commission of the first two items of the signals 値 in the first ordered set and the average of the last two items of the signals 该 in the first ordered set To get the number of the third discriminant index. 9. An image color interpolation method applicable to a photosensitive element having an ax b Bayer image array, the image color interpolation method comprising: randomly selecting an ηχ η pixel array from the ax b Bayer pixel array, wherein A signal 値 is located in a central pixel in the center of the ηχ η pixel array; one above the center pixel, one below, one left, ~ right, one upper left, one lower left, one upper right, and one right are calculated, respectively. The change rate of the signal 値 below, each of which has at least one red light signal 値, at least one green light signal 値, and at least one blue light signal 値; according to the formula T = k, x Min + k2x (Max-Min) Calculate a predetermined threshold 値 T, where Min is the minimum 値 of the signal 値 change rate, Max is the maximum 値 of the signal 値 change rate, and & and k2 are real numbers; select a part of these directions to form a set Where the rate of change of the signals 値 in those directions in the part is less than the predetermined threshold; add the red light signals 値, the green light signals 値, and the blue light signals 内 in the set, respectively, Get a red light signal 値 total G, 33 200536376 12223twf.doc a green light signal 値 total Rsum and a blue light signal 値 total: 8_; and the red light signal 値 total Gsum, the green light signal 値 total, the blue light signal値 Sum Bsum, the red light signals 値, the green light signals 値, and the blue light signals 値 calculate an unknown green light signal 该 in the central pixel. 10. The image color interpolation method as described in item 9 of the scope of patent application, wherein when the signal 値 of the central pixel is a red light signal 値 r5, the formula g5 = 1 * 5 + GRdifr = r5 + (Gsum- U / Sets finds g5, where g5 is the green light signal 该 in the central pixel, and Sets is the number of directions in the set. 11. · The image color interpolation method described in item 9 of the scope of patent application, When the signal 値 of the central pixel is a blu-ray signal 値 b5, use the formula g5 = b5 + GBdifr = b5 + (G picture-Bsum) / Sets to find the green light signal 値 g5 in the central pixel. 12 · 如The image color interpolation method described in item 9 of the scope of the patent application, wherein when the signal of the central pixel is a green light signal meaning, and the two blue light signals are centered μ and ㈣, and the two red light signals are & The heart and the signal of the green light signal 値 Gy above, below, left and right pixels 値, a green light signal 値 & 〃 2 is the signal of the pixel above the blue light signal 値, a green The light signal 値 Gj ^ + 2 is the signal of the pixel below the blue light signal 値, a green light signal Is the signal of the left pixel of the red light signal 値 & seven, the signal of the green light 値 is the signal of the right pixel of the red light signal 値, and the green signals 该 in the ηχ η pixel array When both are known, the method further includes: calculating a difference KR between the red light signal 値 and the green light signal 每一 in each of the pixels having the red light signals ，, where A (x- = κ and A (x +1, less) Second, less-7? χ + 1 > /, 34 200536376 12223twf.doc Some differences: the weights W in these pixels where (G -Π.r + l, > · (Gx , y ~ Gx ~ 2, y and vx-\, y 1 + r 1 i (GX-ly-Gx + ly) 2 _L (Gx, y-GX + 2, y) 2Λ 1 卞 V l 2 J ten l 2 JJ calculates the difference KB between the blue light signal 値 and the green light signal 値 in each of the pixels having the blue light signals 其, and the center (χ, γ =,, less +1, calculates each of the The weight of the rate KB in these pixels corresponds to vx, y- \ 1 + Gx, y ^-° χ ϋless -2 \ 2 and the text, less +1 1 + G less +2. Formula A (㈤, find the center% -1, less + Wx + 1, the difference between fewer pixels (M); use the formula + ~ (· ^ + 1) ×% Out of the central U Wx, “l pixel difference 値, use the difference between the central pixel and the green light signal of the central pixel to define the red light signal 値 L, the center (^ ); And 35 200536376 12223twf.doc use the difference between the central pixel (x, W and the green light signal of the central pixel 値 to infer a blue light signal 値 ^ of one of the central pixels, among which. 13. The image color interpolation method as described in item 9 of the scope of patent application, wherein when the signal 値 of the central pixel is a Blu-ray signal 値 & 〆 and a red-light signal 値 & ^^ is the Blu-ray signal 値The signal 値 in the upper left pixel, a red light signal 値 ^^, ^ is the signal 値 in the lower left pixel of the blue light signal 値, and the red light signal 値 ^^^ is in the upper right pixel of the blue light signal 値The signal 値, a red light signal 値 is the signal 値 in the lower right pixel of the blue light signal 値 L, and the signal 値 in the upper left pixel of the red light signal 値 ^^^ is a green light signal 値, the red The signal 値 in the lower left pixel of the optical signal 値 is a green light signal 値, and the signal 红 in the upper right pixel of the red light signal 値 < +1, μ is a green light signal 値 G + 2, y_2, which The red light signal 値 4 + 1, the signal 下方 in the lower right pixel is a green light signal 値 A + 2, γ2, where the green light signals 该 in the ηχ η pixel array are all known. Including: Calculate the upper left pixel, lower left pixel, upper right pixel, and lower right pixel of the central pixel, respectively. The difference between the green light signals 値 and the red light signals 値 KR 'where A (X- =' KR (x + \, y- \) = Gx ^ Xy_x -Rx + Xy_x 'and ruler + -7 ^ 七 +1 'and +1, less +1) = G ^ + i -1, state; Calculate the weight w of each of these rates 値 KR, where 1 + Gx, y ~ Gx -2, less -2, 2V 36 200536376 12223twf.doc yx + \, y- 1, less +1 vx + \, y + \ 1 + V + l ^ .r + l, v-1 2 r ^ x + \, y- \ ~ G 2 Factory ~ G , ^ Χ + \, γ + \ y 1 (Gx,-~ Gx + 2, y-2 ^ 7 > Ten JV 2 JJ \ 2 _L i L '~ ^ x-2, y + 2 2Λ 卞 J 2 JJ Factory_Factory ^ χ, γ ~ ^ + 2, ^ + 2 and , Less +1 with the equation + A (x + 1, less + l) x% κ (χ ^ _KR (X ^ yl) XWX-l, y-l + KR {X + ly-^) XWX + ly ^ R Wx-ly-l + Wx + l, y- \ Find the difference center of the central pixel (χ, β; and use the difference center of the central pixel (x, W and the green light signal of the central pixel) Reverse the red light signal 値 L of one of the central pixels, where: 14. The image color interpolation method described in item 9 of the scope of patent application, wherein when the signal 该 of the central pixel is a red light signal 値 7 ^ And a blue light signal type ^^ is the signal 値 in the upper left pixel of the red light signal 値, a blue light signal 値 is the signal 値 in the lower left pixel of the red light signal 値, and a blue light signal 値 ^^^ Is the signal in the upper right pixel of the red light signal frame, a blue light signal pattern 0 + 1 is the signal in the lower right pixel of the red light signal frame, and the signal in the upper left pixel of the blue light signal frame値 is a green light signal 値 Gx〇_2, the blue light signal 値 see the signal in the lower left pixel of ^ + 1 値 is a green light signal 値 σγ_2ν + 2, the blue light signal 値 see the upper right pixel of + 1 ^ The signal 値 is a green Signal, the blue light signal (see the signal in the lower right pixel of +1 ,,, +1) is a green light signal (G + 2, p2, 37 200536376 12223twf.doc) When these green light signals 値 are all known, they further include: rf calculates the green light signals 値 and the blue light signals r of the upper left pixel, lower left pixel, upper right pixel, and lower right pixel of the central pixel, respectively. The difference is 値 Κβ, where ~ (X ——, + 1) —— see 'and ~ (χ -1, less +, ㈣-see seven +1, and 4 + 1, less + 1) = G coffee 5 One; Calculate the weight w of each of these rates RKR, where xl, less -1 x + 1, less one 1 and -1, less +1 text +1, less +1 1 + (Q p Chuan, less + l ^ -1. V-1 VV 2 f 1 + ((Z _ generalized -1, less +1 '^ × + Ι, ^-Ι VV 2 f 1 + ^ X + \ yy- \ ~ GX- \ y +] V \ 2 f 1 + ^ x- \ yy- \ ^ x + \, y + \ \ \ 2 \ 2 (Gx ^ Gx ^ y + 2 \ 2 rr Yiguang νΛ Ux, y ~ ^ jc + 2, ^ + 2 2 ( Gx, y ^ Gx_2 ^ (Gx, y-Gx ^ y-2, 2V and Predicate meaning by equation] + Α (χ + 1 ,; μ-1) χ ~ 1Η + Α (χ-1, γ + 1) XM ^ Q + ㈣% points! + WKiss +% + 丨 Find it The difference between the central pixel δ Α (χ, j;); and 38 200536376 12223twf.doc using the difference between the central pixel and the green light signal of the central pixel. ^ Inferring the blue light signal of one of the central pixels. & 15. The image color interpolation method according to item 9 of the scope of the patent application, wherein the ηχ η pixel array includes a 5x5 pixel array. 39