TWI282519B - Color interpolation method with directed weights - Google Patents

Abstract

A color interpolation method with directed weights is disclosed, which provides an algorithm for interpolating pixels into raw data, the method comprising: extracting raw data; generating a digital image of a color filter array; then calculating a weight of a neighbor pixel of a raw pixel, calculating its bilinear interpolation value, and calculating a first set of pixel differential values between a plurality of neighbor pixels and the bilinear interpolation value; and then obtaining a first interpolation pixel and a second interpolation pixel interpolated into the raw pixel, wherein the pixels are interpolated according to each directed weight of the raw pixel, or interpolated according to the weights after determining each directed weights of a color differential value domain applied in the raw pixel; thereby inferring the interpolation pixels of each raw pixel, so as to reduce an edge blur effect generated during a color image recovery.

Description

1282519 IX. Description of the invention: [Technical field of invention] - Color interpolation method with direction and weight, especially to provide an algorithm for calculating the image data of the secret age, to improve the edge blur caused by interpolation The problem. V* [Prior Art] • The digital camera principle uses a lens to project an image onto a Charge Coupled Device (CCD), which is a sensor of a digital camera, by which a camera lens is attached by a charge coupled device. The image taken is converted into a digital image signal, and the digital image signal is stored in the storage medium through the processing of the electronic circuit. However, since the CCD component port can sense the intensity of the light, the color change cannot be felt. When performing digital sampling, a color separation lens must be added in front of the photosensitive element. This color separation filter is a three-color color separation method using RGB three primary color separation method. The values are mixed into a full color image. ® Due to the cost and size of the digital camera, it is generally not considered to use three charge-matching components, but to use a single-charged horse-human component, so that each pixel has only one color of R, G, and B. The gray level of the element, therefore, the conventional technique requires mathematical processing of the result obtained by the photosensitive element to reconstruct the color element lost by each pixel. The conventional image interpolation method includes two types: The first type is fixed image interpolation, and the adjacent points of the color elements left by such interpolation are fixed, and the interpolation method itself does not detect edges. The ability of 12825.19, thus causing edge blurring, such interpolation methods include nearest interpolation, bilinear interpolation, and smooth hue transition interpolation. The image is partially blurred and the image is blurred. The second type is non-stationary image interpolation. The interpolation method is not fixed by the interpolation of the color elements. The interpolation method itself has the ability to detect edges, so it can be reduced. Edge blurring, such interpolation has a first edge detection interpolation method, a second edge debt interpolation method, and a linear interpolation with Laplacian second-order correction linear interpolation Interpolation, variable number of gradients interpolation, pattern recognition interpolation, the interpolated image can effectively reduce the edge blur in the horizontal and vertical edges. Phenomenon, but the edge passing through both the 45-degree angle and the 135-degree angle will still have blurred image edges. The image captured by a single coupling element is represented by a Color Filter Array (CFA). In the color filter array, only one pixel (pixel) is used for each array position, and the other two are left. Pixels must be filled by interpolation to restore the image. In the prior art, in the photosensitive element such as CCD or CMOS, each point of the color filter and the light film array individually receives one of red, green and blue, and each color can be divided into 256 levels according to the depth and depth. As shown in the first figure, a partial view of a color filter array includes a plurality of pixels, a pixel at the beginning of B is a blue pixel, a pixel at the beginning of r is a red pixel, and a code beginning with a G is a green pixel. The missing pixels must be reconstructed by interpolation 1272519. The Republic of China patent TW548956 exposes a method for judging the weight of the cross weighting judgment method. As shown in the second figure, it is a partial facet of an image, and the boundary between black and white (indicated by different stripes) is the edge of the image, which passes through the facet in the vertical direction: the face is divided into the first area 21 With the second area 22. The color filter array can be shown in the second β diagram, wherein the second map is divided into two regions: the first region 21 is red, green, and blue, and the second region 22 is red, green, and The neighboring pixels of the array color pixel B7 having the blue value are green pixels G3, G6, G8, and G11, etc. = the weight of the adjacent four pixels and the fairy-color pixel G8 is determined by the weight value W3, the green pixel The weight of GU is as follows: W2= 1 = 0.003f W4 ^l+(200-〇f+ (0^0)2 〇β〇05 ^1+(0-200^+(0-200^ ο)2· .) 2-1 The pixel to be interpolated is adjacent to each pixel as a weight ratio and "the pixels used in the '10I$' pass through the pixel to be interpolated (this f:, ,, B7)' The cross weight judgment method. After the weight value is calculated, it is found that W1 is equal to W4, #u + y and both are much larger than W2 and W3, so it can be judged that interpolation is performed at this time, and G3 and G11 are interpolated, but in fact, For the plurality of categories to be inserted, in terms of categories, it is the same as G3, G6 and G11 (the same as the first-region 21) '❿G8 belongs to the other-class, and the cross weight judgment method will be G6 (weight 1272519 · W2) made a wrong classification. The second A picture shows An image edge passes through the facet in an oblique manner, and includes a third region 31 having red, green, and blue values of 2 turns and a fourth region 32 having red, green, and black values of 0. The third B The plan is a schematic diagram of the color filter array of the third B diagram. In the figure, the adjacent pixels of the blue pixel B7 have green pixels G3, G6, G8, and G11, etc., and the G7 is interpolated to obtain the weight, and the weight is also used. Determining the correlation between the neighboring four pixels and the position to be interpolated, such as the weight of the green pixel G3, the weight of the green pixel G6, the weight of the green pixel G8, and the weight of the green pixel G11. W4, the formula is as follows

Wl=_-_- _1 ♦ (4)—When G11 is printed—W3— ^HGSr-Gef +(B7-B5y ^+(0-200^ +(200^200^ — 〇服^4^+(G6-GSfίl ·(BΊ--B9f ^1+(200-0)2 +βΟΟ^〇γ〇.〇〇3w

Vr+(G3-Gll| +(B7-B12j ^1+(200-0)2+(2^T〇f_0.〇〇35 After the above weights are calculated, the weight W1 is equal to W2 is 〇·005 W3 is equal to W0 is 0·0035, and the weights are similar. Therefore, the cross-weight judgment method will cause such image edge blurring effect. Therefore, when the image edge passes the 45-degree angle to the left, the weight judgment method cannot be correct. Judging by the technique, the interpolation method can not effectively solve the phenomenon that the interpolated image has edge nucleus paste, because the interpolation method itself cannot lie sensitively to the position where the edge passes, so it is impossible to In view of the above, the present invention proposes a color (four) interpolation method with directional weights. [Invention] A color interpolation method with directional weights is used to solve the conventional image interpolation. The method still causes the edge to appear (4), providing - (4) the original 1222519 original pixels, and the method of generating the first pixel of the complex pixel: the weight of the adjacent pixel direction, the calculation of the original pixel 盥 - ^_ interpolation Calculating a plurality of adjacent pixels of the original pixel and interpolating the original pixel =, and Prime difference, then pay near the pixel direction weight, calculate Park Shu's first-group = value and calculate the second interpolation of the original pixel: the value of the cattle ^ ^ and $ calculate the original pixel's first-group bilinear The interpolated eigenvalue field performs image interpolation, and the color difference-value domain is evaluated by a bilinear interpolation method. The above-mentioned first interpolated pixel is substituted by a first group of adjacent pixel directions 苐 a group of pixel difference values. The following equations are given: ', Y, i=Xi, i+(Wi-l, iX Ki-,, J + Wi, MX L.mH-Whmx Ki+1, j + Wi j + l original pixels in - color The filter, the light film array (cta) towel is the first interpolated pixel of the original image of the county pixel, and the second interpolated pixel is substituted by the second set of adjacent pixel direction weights, a set of pixel difference values into the following equation inferred:

Z^KW^wx + Ki+1„1 + w1+1,,lX K_+u+1 + Wm,J+1X Km, ,ι)/(Ψι_1; m + Wi+1 H + Wi+1 j+ i + Wi ^J+1} ? =1 and j are the positions of the plurality of original pixels in a color filter array FA), and Xi j is the original pixel, which is the interpolation of the original 1282519 * pixels The second embodiment of the present invention uses the following steps to obtain the interpolated pixels of the original planar pixels: μ ° first captures the original pixels, and generates a digital image of the array of color pupils. The first group of pixels is the weight of the neighboring pixel direction, and the first group of pixel difference values of the adjacent pixels of the first-original pixel is obtained as the inner pixel = the first pixel of the first impurity pixel, and then the first original = ft group is calculated. The weight of the adjacent pixel direction, the first pixel of the first pixel are adjacent to the group pixel difference value to obtain a second interpolation pixel interpolated to one of the first-origin pixels. The weight of the pixel direction, the first set of pixel difference values of the adjacent pixels, to obtain the second interpolation pixel; and calculate the second group of the second original pixel like the symplectic ##方, weight, the first a second pixel of each of the adjacent pixels of the original pixel, and a pixel value of the second pixel is transmitted to the second pixel. The value of the first pixel is within the weight of the first group of adjacent pixels. Inserting the first set of pixel difference values of the near pixel to obtain the __ pixel of ri silly ri; and calculating the third original two standbys, the weight of a set of adjacent pixel directions, and the third original pixel adjacent to each other The first group of pixel difference values, and finally the second interpolation pixel of the third original pixel is obtained. [Embodiment] A single-charge coupled device (CCD) is used as a photosensitive device in the digital camera. The image is represented by color filter, light film array (CFA), 1282519. In the color filter array, only one pixel is taken at each array position. The other two missing pixels must be interpolated to fill the restored image. In the prior art, the image pixel misclassification or the image edge blurring effect caused by the cross weighting judgment method is used, so the present invention proposes a non-cross weight judgment method for reducing edge blur caused by color image restoration. With The color interpolation method of the directional weight. The invention includes two main features: the first is the original directional weight for the weight judgment, and the second is the color difference value field for the weight judgment. For an embodiment for weight determination, please refer to the schematic diagram of each adjacent pixel in FIG. 4A, in which the image edge passes through the facet in the vertical direction. The left edge region 41 is red (r), green (g) and blue. (B) The value is the area of 〇, and the edge right area 42 is the area where the red (R), green (G), and blue (B) values are 2 。. The blue pixel B7 is required for this embodiment. The interpolated pixel has adjacent pixels as green pixels G3, G6, G8 and G11. In this example, if the missing color, such as red and green, is to be interpolated near the blue pixel B7, the weight ratio of each adjacent pixel needs to be calculated first. Then, the correlation between the neighboring pixels and the interpolation position is determined. For example, the weight of the green pixel G3 is W2, the weight of the green pixel G6 is W2, the weight of the green pixel G8 is W3, and the weight of the green pixel G11 is W4. . Wherein the embodiment discards the conventional technique, the neighboring pixels across the interpolated pixel (in this case, B7) are used to calculate the weight, and the fourth B, the fourth C, and the fourth D are the fourth. The pixel B7 shown in the figure E and the adjacent pixel _ are weighted by a plurality of pixel positions passing in the direction of the arrow, and the calculation formula is as follows. For example, the fourth B-, wherein the weight W1 of the green pixel G3 is ·· wi - 7 1 _ 1 —] a/i+(G3-gi)2+(B7-B2)2 12825.19 The weight W2 of the color pixel Gg is :

Vr+(〇-〇)2+(〇^〇F=1 is ··· W3 ^ Tens, the weights W1, W2 and W4 are the weights = _, and the pixels G3, G6 and G11 have the same weight The value, ^ direction element is also in the same area, so the hair error proposed by the hair ride, St color _ method to get the correct classification, improve the knowledge of the wrong and enter the corner at the edge of the image at an oblique angle (such as 45 degrees) In the embodiment of the screen, as shown in FIG. 5, the color filter and the light film array, wherein the screen is divided into an upper area 51 and a lower area 52, and the upper area 51 is red (r), green (G) and blue ( B) The value is 2 〇〇, and the area 52 below the edge is the area where the red (R), green (G) and blue (B) values are ◦. The blue pixel B7 needs to be interpolated in this embodiment. Pixels whose neighboring pixels are green pixels G3, G6, G8 and G11. To interpolate the missing color near the blue pixel B7, the weight ratio of each adjacent pixel needs to be calculated, and then the correlation between the adjacent pixels and the interpolation position is determined by the result. For example, the weight of the green pixel G3 is the weight of the green pixel G6, the weight of the green pixel G8 is W3, and the weight of the green pixel G11. W4. The embodiment discards the prior art by using the cross weight judging method, and calculates the weight of each neighboring pixel by the relationship between the pixel B7 and the neighboring pixels as shown in the fourth to fourth Eth diagrams above, and passes through the direction of the arrow. The multiple pixel positions are weighted and the calculation is as follows. 12 =11282519

Wl= W2: W3= The weight W1 of the green pixel G3 is: — 1 ___1_ ^l+(Gl-G3f +(B7-B2f ^1+(200-200)2 +(200-200/where the green pixel G6 The weight W2 is: · - \ __1_ ^1+(G4-G6)? +(B7-^B57 /l+(200-200^ +(200-200]? Where the weight of the green pixel G8 is W3: =1 = 0.005 ^(GIO-GS)2 +(B7^B9f ^1+(0-0)2+(200-Ο)2 where the green image is the weight of W1 of G11 is ··^ Ji+(〇-〇) 2+(200-0)2, after the weight calculation, the weights W1 and W2 are 1, and the weights W3 and W4 are 〇·〇〇5, and the pixels G3 and G6 have the same weight, actually The two pixels are also in the same area, and the pixels G8 and GU have the same weights. The two pixels are actually in the same area. Therefore, the color interpolation method with directional weights proposed by the present invention is solved. The prior art has the lack of oblique edge blurring phenomenon. The color interpolation method with directional weight proposed by the present invention can also perform image interpolation by using the shape difference value domain, and apply the color difference of each adjacent pixel to the domain. The judgment of the weight, in order to improve the quality of the image after interpolation, also = the value of the H-color difference The first embodiment of the present invention is implemented in a weighted manner on red, green, and blue = RGB space, wherein the directional color interpolation method of the present invention includes The following multiple steps: 1. Interpolating the red pixel (R) and the green color pixel (G) at the position of the blue pixel (8) to generate an interpolated value of the green pixel: a color filter array as shown in FIG. Referring to the first figure, W4: = 0.005 13 1282519 shows the position of each pixel of the array, for example, by interpolating the green pixel (G44) on the red pixel R44, first calculating the green pixel of the red pixel R44 adjacent to each direction ( G) Weight, such as the following equation: Relative to the red pixel R44, the weight W34 of the adjacent green pixel 〇 34 direction is 1 _ n _ —> · /l + (R44 - G14)2, red pixel R44 The weight w43 of the direction of the adjacent green pixel G43 is —— ", , · φ VT+ (R44 -^42)^ (G43 - G41)2 'The red pixel R44 is adjacent to the green 彳i G54 direction weight 14 is φ + (R44 - R64)2 + (G54 - G74)2 , green pixel G adjacent to red pixel R44 The weight w45 of the 45 direction is "j- 丨丨丨·丨丨·_ — · VI + (R44 - R464)2 + (G45 - G47)2, k The above denominator adds 1 to the root number to avoid The denominator is a question of ambiguity, but it does not affect the judgment of weight. Then calculate the interpolated value produced by the bilinear interpolation method. In this case, the color gamut difference value range is evaluated by this bilinear interpolation method, that is, the average value of two adjacent homochromatic systems is calculated, such as red pixels. The bilinear interpolation of R44 and R24 is the average value (R44+R24)/2, the bilinear interpolation of R44 and R42 is (R44+R42)/2, and the bilinear interpolation of R44 and R64 is ( R44+R64)/2, the bilinear interpolation of R44 and R46 is (R44+R46)/2. Further calculating the pixel difference between the adjacent pixels (such as G34, G43, G54, G45, etc.) and the bilinear interpolation value, such as K34=G34-(R44+R24)/2 is the pixel difference between the red pixels R44 and R24. 'Other pixel difference values such as K43=G43-(R44+R42)/2; K54=G54-(R44+R64)/2; K45=G45-(R44+R46)/2 and so on. The above-mentioned adjacent pixel-to-pixel difference values K34, K43, K54, and K45 are multiplied by the weights of 1282519*values W34, W43, Ws4, and W45' to calculate the interpolated green pixel (G44) on the red pixel R44, as in equation (1). Said: G44=R44+(W34X K34 + W43X K43 + W54X K54 + W45X K45)/(W34 + W43+W54+W45)--------________________________ (1) However, 'interpolated blue pixel (B) The green pixel (G) step is like the green pixel (G) in which the red pixel (R) is interpolated as described above. 2. Interpolating the blue pixel (B) at the red pixel (R) position and the red pixel (R) at the blue pixel (B); as shown in the sixth figure, 'in the color filter array For example, the blue pixel (B44) is interpolated on the red pixel R44. The weight of the red pixel R44 adjacent to each blue pixel (B) direction is calculated first, as shown in the following formula: Relative to the red pixel R44, adjacent to the blue pixel B33 The weight W33 of the direction is 1 - · a / 1 + (R44 - R22) 2 + (B33 - B11) 2, and the weight W53 of the direction adjacent to the blue pixel B53 is ~r _ 1 _ with respect to the red pixel R44.彳1 + (R44 - R62)2 + (B53 - B71)2 ' With respect to the red pixel R44, the weight W55 in the direction of the adjacent blue pixel B55 is 1 1 :: -. · VI + (R44 ~ R66)2 + (B55 - B77)2 , relative to the red pixel R44, the weight W35 in the direction of the adjacent blue pixel B35 is ~f=-1 - 〇+ (R44 - R26)2 + (B35 - B17)2 After the weight (W) in the angular direction, the value of the interpolated blue pixel B44 is obtained from the green pixel G44 obtained in the previous step and the red pixel R44 to be interpolated. The green pixels G33, G53, G55, and G35 are interpolated in each direction obtained in the first step, thereby further calculating the red pixel R44 to be interpolated and the adjacent pixels (such as B33, B53, B55, B35, etc.) to be interpolated. The pixel difference value (K), such as the red pixel R44 and its oblique angle adjacent to the image 15 1282519, such as K53=G53-B53, the pixel difference K33=G33-B33, K55=G55-B55, K35=G35-B35, etc. . Next, the relationship between the adjacent pixel difference values K33, k values w33, W53, W55, w35 (such as equation (2) ^ R44 is interpolated with blue pixels (B44), where the equation value is referenced and substituted into the equation (2) ): K35)/(W33 one--(2)

B44=G44-(W33x K33 + W53X K53 + W55X K55 + W35X + W53 +W55 +W35)--------------------

In this step, the red pixel on the blue pixel (B) is interpolated (the step is as the blue pixel (B) of the above-mentioned interpolated red pixel (R). 3. The red at the position where the green pixel (G) is interpolated. The pixel (R) and the blue pixel (B): m is an example in which the red pixel R45 is interpolated on the green pixel G45 in the color filter array shown in FIG. 7 , and the green pixel G45 is first calculated in all directions. The weight, as shown in the following formula: Relative to the green pixel G45, the weight W44 of the direction adjacent to the red pixel R44 is 1 - Γ; λ / 1 + (G45 - G43) 2 + (R44 - R42) 2 relative to the green pixel G45 The weight W55 of the direction adjacent to the blue pixel Β55 is -;; V1 + (G45 - G65)2 + (B55 - B75)2 is relative to the green image ^ G45, and the weight W46 of the direction adjacent to the red pixel R46 is ~j= ===:=:= •.1; VI + (G45 - G47)2 + (R46 - R48)2 Relative to the green pixel G45, the weight W35 in the direction of the adjacent blue pixel B35 is ~Γ==^— -0 VI + (G45 - G25)2 + (B35 - B15)2 After obtaining the above-mentioned weights in the respective directions, the interpolated values of the pixels adjacent to the green pixel G45 are obtained as a result of the first step, such as Red! 282519 · Green pixel G44 of pixel R44, green pixel G55 of interpolated blue pixel B45, green pixel G46 of interpolated red pixel R46 and green pixel G35 of interpolated blue pixel B35. Result of step 2 Knowing the interpolated values of each adjacent pixel, such as • the blue pixel B44 of the red pixel R44 and the blue pixel B46 of the interpolated red pixel R46, and the pixel difference (K) of the neighboring pixels B35 and B55, such as K44=G44-B44, K55=G55-B55, K46=G46-B46 and K35=G35-B35, etc. The present embodiment can be obtained from the above-mentioned weight value and pixel difference. • Blue pixel of the interpolated green pixel G45 B45, as in equation (3): B45=G45-(W44x K44 + W55X K55 + W46X K46 + W35X K35)/(W33 + W53 + W55 + W35)--------------- -------------- (3) - Also, the step of interpolating the red pixel R45 of the green pixel G45 is the same as the above-mentioned interpolated blue pixel B45. ^ 4. Reconstructing the image green pixel Color correction: After steps one, two, and three, the color of the green pixel left in the red plane position needs to be corrected, but this is a selective step, not a must Because of the first step, the color difference range is evaluated by bilinear interpolation. This evaluation method will cause interpolation error, so in this final step, it can be selectively interpolated. Correcting this error, this calibration step will increase the image signal-to-noise ratio (Peak Signal—1〇—N〇ice

Ratio, PSNR) is about 2-4 decibels, for example, to correct the green pixel G44: as in step 1, the adjacent red pixel has weights in each direction: ^/l + (R44 - R24)j^(G34 - G14)2, λ/1 + (R44 - R42)j2T(G43 - G41)2, yjl + (R44 - R64) 2^7g54 - G74)2, 17 1 l282519 φ + (R44 - R46)2 + (G45 - G47Y : The difference is that the calculation method of the difference Κ34, K43, K54, - between adjacent pixels is as follows: Step 3 shows that the red pixel (R) at the position of the green pixel (G) is interpolated. Bilinear interpolation method to reduce the error. Therefore, the difference of each pixel is K34=G34-R34, K43=G43-R43, K45=G45-R45, etc. Finally, the weights W34, W43, W54, W45 corrects the green pixel G44 of the red pixel R44, as in Equation (i) · G44 = R44+ (W34X K34+W43X K43 + W54X K54+W45X K45)/(W34 + W43+ W54+ W45)----- ------------------------ (1) Other green pixel color corrections that are left at the 1-color plane position are similar to the above-mentioned corrections in the red plane position. Falling green pixels., * As described above, the first embodiment of the present invention The weight of the original red pixel R is adjacent to each other (such as vertical, horizontal, oblique direction, etc.) and the pixel difference, and the interpolated values G and Β of the pixel are given, and the blue pixel 推 = interpolated pixel R is similarly derived. And G, and then calculate the interpolated values R and Β of the green pixel G, 10 卩 to obtain the _ value of each pixel of the original original flat towel. The second embodiment of the present invention is different from the first embodiment. In the plane, the pixels are interpolated according to the weights of the pixels in each county, and the weights of the pixels in the different values of the color are different, and the pixels are adjacent to each direction (such as vertical, horizontal, oblique, etc.). The horse ki, the pixel difference, the interpolated value of the original pixel is obtained to obtain the complete interpolation formula of each pixel in the mouth plane. In the second embodiment, the weighting method is implemented in the green-red color The color value interpolation method with the directional weight value of 1282519 proposed by the present invention includes the following multiple steps: 1. Interpolating red pixel (R) and blue pixel (B) ) a green pixel (G) at the location to generate an interpolation of green pixels For example, in the color filter array shown in FIG. 6, the green pixel (G44) is interpolated on the red pixel R44 as an example. First, the green pixel (G) weight of the red pixel R44 adjacent to each direction is calculated, as shown in the following equation: With respect to the red pixel R44, the weight W34 of the direction of the adjacent green pixel G34 is ___ 1 Τ ΐ ΐ 44 44 R R R 24) 2 + (G34^^ ; the weight W43 of the direction of the green pixel G43 adjacent to the red pixel R44i is a/ T+7r44 - R42)2 + (G43^G41)2, the weight of the red pixel R44f near the green pixel G54—for λ[ι~+ (R44 - R64)2 + (G5i^T(j74)2, The weight ψ45 in the direction of the green pixel G45 adjacent to the red pixel R44i is Vl + (R44 - R464) 2 + (G4^Tg47) 2, and the above denominator adds 1 in the root number to avoid the problem that the denominator is 〇, Does not affect the judgment of weights. Recalculate the pixel difference between adjacent pixels (such as G34, G43, G54, G45, etc.) of R44 and the bilinear interpolation value (as described in the first embodiment), such as: Ks4=G34-(R44+R24)/2 K43=G43-(R44+R42)/2 K54=G54-(R44+R64)/2 K45=G45-(R44+R46)/2 By the difference Km, K43, Km, Κα between the adjacent pixels The weights W34, 19 1282519 W43, W54, W45 calculate the interpolated green pixel (G44) on the red pixel r44, as described in the equation (丨) of the first embodiment: G44=R44+(Ws4X K34 + W43X K43 + W54X K54 + W45X K45)/(W34 + W43 + W54 + W45)--------------------------------(1) Same as ' The step of interpolating the green pixel (g) on the blue pixel (B) is as the green pixel (G) in which the red pixel (R) is interpolated as described above. 2. Interpolating the blue pixel (β) at the red pixel (R) position and the red pixel (R) on the blue pixel (Β); as shown in the sixth figure, in the color filter array For example, the blue pixel (B44) is interpolated on the red pixel R44, and the weight of the red pixel to the direction of each blue pixel (B) is calculated first. Different from the first embodiment, the original pixel is used to calculate the weight of each direction, and the second embodiment is a green (G)-red (R) color difference value or green (G)-blue (β). The color difference value calculates the weight of each direction to obtain an accurate interpolated value. First calculate the color difference value of each pixel on the original pixel plane, such as the green-red color difference value or the green-blue color difference value, as shown in the first oblique direction direction 81 shown in the eighth figure, the color difference of each pixel in the direction Value

Kr44=G44-R44, Kr22=G22-R22, Kb33=G33-Β33 and

Kbll=Gll-Bll, etc.; and the color difference value of each pixel in the second oblique direction 82 is Kr44=G44-R44, Kr62=G62-B62, Kb53=G53-B53 and Kb71=G71-B71, etc.; third oblique angle The color difference value of each pixel in the direction 83 is Kr44=G44-R44, Kr66=G66-R66, Kb55=G55-B55 and Kb77=G77-B77, etc.; the color difference value of each pixel in the fourth oblique direction 84 is Kr44=G44 -R44, Kr26=G26-R26, Kb35 two G35-B35 and 20 1282519

Kbl7=G17-B17 and so on. From the color difference values in the above directions, the weight W33 of the direction adjacent to the blue pixel B33 with respect to the red pixel R44' is ▽1 + (Kr44 - Kr22^T(Kb33 - Kbl 1)^, relative to the red Pixel R44, the weight W53 of the direction adjacent to the blue pixel B53 is ~7 --=======-- 1 ........"; VI + (Kr44 - Kr62)2 + (Kb53 - Kbl\)2 Relative to the red pixel R44, the weight W55 of the direction adjacent to the blue pixel B55 is ~r 1 · V1 + (Kr44~Kr66)2+(Kb55-Kb77)2 relative to the red pixel R44, adjacent to the blue The weight W35 in the direction of the pixel B35 is ~1 - ———— =-〇VI + (Kr44 - Kr26)2 + (Kb35 - Kbl7)2 After learning the weight (W) of each of the above-mentioned oblique directions, The pixel difference between the adjacent pixels of the oblique angle is K33=G33-B33, K53=G53-B53, K55=G55-B55, Km-G35-B35, and the green pixel G44 obtained in the previous step and the red to be interpolated Pixel R44 to get the value of the interpolated blue pixel B44, as in equation (2): B44=G44-(W33X K33 + W53X K53 + W55X K55 + W35X K35)/(W33 + W53 + W55 + W35)---- ---------------------------- (2) And the same reason can be calculated to interpolate the red pixel R44 and bevel Near pixels (such as B33, B53, B55, B35, etc.). In this step, the red pixel (R) on the blue pixel (B) is interpolated as the blue pixel (B) in which the red pixel (R) is interpolated. 2. Interpolating the red pixel (R) and the blue pixel (B) at the green pixel (G) position: ^ ^ Interpolating the red pixel G45 in the color filter array shown in the seventh figure For example, the pixel R45 needs to calculate the weight of the green pixel G45 adjacent to each of the 21 1282519 directions, and the weight is calculated by using the color difference value in each direction to obtain an accurate interpolation value, as shown in the following formula: Relative to the green pixel G45 ' Its weight adjacent to the direction of the red pixel R44 is 5 W "G43"; + (Κτ44^Γ, where the color difference value is Kr44=G44-R44, Kr42=G42-R42; relative to the green pixel G45, adjacent to the blue pixel B55 The weight of the direction is ^55 ^ V^+(G45 -G65)2 + (Kb55 -Kb75)2, where the color difference value is Kb55=G55-B55, Kb75=G75-B75; k is relative to the green pixel G45, adjacent The weight 趴6 of the red pixel R46 direction is 7^^-〇47)2 + (&46-;^48)2, where the color difference value is

Kr46=G46-R46, Kr48=G48-R48; with respect to the green pixel G45, the weight of the direction adjacent to the blue pixel B35 is λ/Ϊ+(G45 - G25)2 + (Kb35 - Kbl5)2, where the color The difference values are Kb35=G35-B35 and Kbl5=G15-B15. After obtaining the weight values of the above-mentioned directions, the interpolated values of the pixels adjacent to the green pixel G45 are obtained as a result of the first step of the embodiment, such as inserting the green pixel G44 of the red pixel R44, interpolating blue. The green pixel G55 of the pixel B45, the green pixel G46 of the red pixel R46, and the green pixel G35 of the blue pixel B35 are interpolated. Then, the interpolation result of each neighboring pixel is obtained by the result obtained in the second step, for example, the blue pixel B44 of the red pixel R44 and the blue pixel B46 of the red pixel R46 are interpolated, and the pixel difference between the neighboring pixels B35 and B55 is obtained. (K), such as K44=G44-B44, K55=G55-B55, K46=G46-B46 and K35 two G35-B35. And the blue pixel Β45 of the green pixel G45 is inserted in the embodiment according to the weight value and the pixel difference value as described above, as in the case of the first embodiment 22 1282519 · Program (3): B45=G45-(W44x K44 + W55X κ, + wx K. + Wx KsO/CWss + W53+ W55 + W35)------~1—(3) The steps of interpolating the red pixel of the green pixel G45 and the process pixel R45 are as described above. Blue pixel B45. 4. Reconstructing the image green pixel color correction: After the steps -, 2, and 3 of the embodiment, the color of the green pixel in the red plane position needs to be corrected, but this is a selective step, It is not necessary to perform, because in step 1, the color difference range is evaluated by bilinear interpolation, j: the evaluation method will cause interpolation error, so in this last step, you can choose Correctly interpolate • This error, such as the correction of green pixel G44 as an example: • Refer to Figure 8 'First calculate the color difference between the original pixel (such as R44) and the adjacent pixel, using the color in each direction The difference value produces weights in all directions adjacent to the red pixel R44: I - -1 - ^ - , ^/1 + (Kr44 - Kr24)2 + (Kr34 - Krl4)2 10 where Kr44=G44-R44, Kr24=G24 -R24, Kr34=G34-R34 and Krl4=G14-R14; W43 = V- , yi + (Kr44 - Kr42)2 + (Kr43 - Kr41)2 where Kr44=G44-R44, Kr42=G42-R42, Kr43= G43-R43 and Kr41=G41-R41; W54= / ,,▽1 + (Kr44 - Kr64)2 + (Kr54 - Kr74)2 where Kr44=G44-R44, Kr6 4=G64-R64, Kr54=G54-R54 and Kr74=G74-R74; W45= - - , ψ + (Kr44 - Kr46)2 + (Kr45 - Kr47)2 23 1282519 where Kr44=G44-R44, Kr24=G24 -R24, Kr34=G34-R34 and Krl4=G14-R14. Then calculate the difference between each adjacent pixel K34, K43, K54, K45, such as

KwG34-R34, K43=G43-R43, K54=G54-R54, K45=G45-R45, and the like. Finally, with the weight of each direction 14,? 43. Boundary 54, 145 corrects the interpolated red • pixel with green pixel G44 of 4, as in step 1 of equation (1): G44=R44+(Ws4X K34 + W43X K43 + W54X K54 + W45X K45)/(Ws4 ^ +W43 + W54 + W45)--------------------------------(1) Other green pixels left in the blue plane position The color correction is also similar to the green pixel corrected for the red plane position described above. The above calculation calculates the weight of the original red pixel R adjacent to each direction (such as vertical, water 'flat, oblique direction, etc.) and the pixel difference value, and the weight value can be calculated as the difference between adjacent pixels in each direction of the original pixel ( K) It is concluded that the interpolated values G and B of the pixel can also be obtained by the difference between the adjacent pixels (Kr or Kb) of the nose in the difference of the color I in each direction, and the blue pixel B is derived in the same manner. The pixels R and G are interpolated, and the interpolated values R and b of the green pixel g are calculated, that is, the interpolation formula of each pixel in the original plane is obtained. The interpolated value of the pixel is interpolated in the relationship between the original pixel and the adjacent pixel, and the step is as shown in the ninth figure: At the beginning, the original image data is captured from the photosensitive element and the color filter is generated. The array (CFA) is a digital image of the Bell pattern 'and the associated pixel position, including a plurality of original pixels to be interpolated pixels, such as the first original pixel, the second original pixel and the third original t pixel' can be respectively represented The red pixel R, the blue pixel B, the green pixel G, and the like, such as the red pixel R44, the blue pixel B33, and the green image 24 1282519 * G45, etc. in the sixth figure, calculate the other two insertions of the original pixel individually. a pixel (step S901); first, first calculating a weight of a first set of adjacent pixel directions of the first original pixel, and an embodiment thereof may calculate each of the neighboring pixels in a vertical or horizontal direction of the first original pixel a weight of a neighboring pixel, such as a plurality of weights w of green pixels such as G34, G43, G54, G45 adjacent to the red pixel R44 in the first step (step S903);

In the embodiment of the present invention, the image interpolation is performed by using the color difference value domain, and the color difference value domain is evaluated by bilinear interpolation, that is, the first original pixel is calculated by the bilinear interpolation method. The average of the same color pixels (the first set of bilinear interpolation values), calculate the bilinear interpolation (R44 + R24)/2 of the original pixels R44 and R24 as shown in the sixth figure (step S9〇5); Crossing 'calculates the first-outer buckle (9) er ~ 郇 1 豕 I 步骤 step S8 〇 5 ^ - group bilinear interpolation value between the first group of pixel difference, that is, the multiple K values in the sixth figure (step S907);

...using the weight (w) and the pixel difference (κ) of the neighboring pixels in the above directions to be interpolated to the pixel of the first-original pixel, that is, to enter the equation, and to be the position of the pixel in the array. , Xi j is the first original two, i, i is the first interpolated pixel of the first original pixel. For example, in the sixth picture, the green image of the original red pixel R44*(10) and its neighboring pixels are Zengshangan', such as the green pixel G33 of the blue pixel B33, and can be interpolated with G53 and G35 in the vicinity of the S9& G55, etc. (Step-Xi.-+(Ww, jX Κμ,, + W.wx Κ, Μ + WiH.-x K1+1, J + Wl, J+ij J j + Wi, ji + Wi+i, j +Wi, j+i)--------(4) 25 1282519 · The right ί4 counts the first—the second of the original pixels _ near the pixel direction Γ Γ 借 第 第 原始 原始 原始 原始Or the weight of each neighboring pixel between the horizontal and the irh prime, as shown in the sixth figure—= near, = color pixel R44, B33, B53, B55, B35, etc. f (step S911); , after 'subtracting each _pixel obtained by the silk type (4) from a plurality of original pixels adjacent to the first pixel, and obtaining a second set of pixel difference values of the first-original pixel #-β pixel (K), as shown in FIG. 6, the pixel difference K33=G33-Β33 of 33, etc. (step S913); using the weight of the second group of adjacent pixels of the first original pixel and the second The group pixel difference 〇〇 is substituted into equation (5), and the interpolation is obtained. a second interpolated pixel of the first pixel, wherein i and j are pixels in the array, the position, Xu is the first original pixel, and Zi j is the second interpolated pixel interpolating the first original pixel. Inserting the blue pixel B44 of the original red pixel R44, the interpolated value of the adjacent pixel, such as the red pixel r33 of the interpolated blue pixel B33, interpolating the milk 3, R35 and R55 with other neighboring pixels (step S915) ;

Zi^-Xi^^CWi-uMX Ki~i, M + WiH, MX KiH, M + Wi + i, j+iX KiH, jH + Wi-1, jHX Ki-1, jH )/(Wi-l, M + WiH, M + WiH, jH + Wi-1, j + 〇-------(5) Finally 'correct the color of the fall (step S917), such as reconstructing the image green pixel. From the above steps Calculating the first interpolated pixel and the second interpolated pixel of the first original pixel, and similarly obtaining the first interpolated pixel and the second interpolated pixel of the second original pixel, and then different from the first The original pixel and the second 26 1282519 · the original original pixel of the original element is its main original pixel, and the other two _ pixels are calculated according to the equation for deriving the calculated interpolation pixel, and all the originals on all the original planes can be derived. Interpolating the pixels. In addition, the weight calculations in the above steps S903 and S911 are performed on the original planes of red, green, and blue as described in the practical example, and the calculation is performed according to the pixel values of the original = pixel directions. Value, and in the second embodiment, the color ¥ difference value field is used for the weight determination of each direction of the original pixel to calculate the pixel difference value, and then further calculate the interpolation value, the main The steps are as shown in the tenth figure: Raw data is first taken from the original pixel plane (step S10), and then the weight of the first set of adjacent pixels of the first original pixel is calculated as 4卩W value ( Step S12), recalculating the first set of bilinear interpolation values of the first original pixel and its adjacent homochromatic pixels (step S14); ^, calculating the f between the first original pixel and the first set of bilinear interpolation values a set of pixel difference values, that is, a K value (step (10), and thereby obtaining a first interpolated pixel interpolated to the first original pixel (step S18); > a second interpolated pixel, and the first The difference between the embodiments is that the color difference value between the original pixel and the adjacent original pixel is calculated first, such as Kr or Kb ^ (step S2〇), and then the second set of adjacent pixel directions of the first original pixel is calculated according to the color difference value. Weight (step S22), recalculating the second set of pixel difference values of each adjacent pixel of the first original pixel, where ^ (step, S24) is obtained by interpolating the second pixel of the first original pixel (step S26) In order to correct the error that may occur in the above steps, the first L initial image can be continuously calculated. ίThe color difference value of the pixel adjacent to i, such as the value (step, hunting the difference between the pixels in the new leaf, to correct the color of the fall (step 27 1282519. S30), such as the process shown in Figure 11 Applying the flow described in the above ninth and tenth drawings to calculate the interpolation value of each original pixel on the original plane, the flow is as follows: · At the beginning, the original image data is captured, and a color filter array (CFA) is generated. a digital image of the Bell pattern (step si〇〇); then calculating a first set of adjacent pixel directions of the first original pixel (step S101);

Recalculating the first set of pixel difference values of the first original pixel, wherein the pixel difference is calculated by using a bilinear interpolation between the first original pixel and its adjacent homochromatic pixels (step S103); " Substituting the weight value and the pixel difference value into the interpolation formula (such as equation (4) or (5) of the present invention), the first interpolation interpolated into the first original pixel is obtained (step S105); - second pixel of the original pixel_near pixel direction

Value (step S1G7) 'calculating the second set of pixel difference values of each adjacent pixel of the first-original pixel (step S1G9) 'Substituting the weight value and the pixel difference value into the interpolation formula' and deriving the interpolation into the first-origin pixel The second interpolated pixel (step Sill); the first set of adjacent pixels of the original pixel can be simultaneously calculated by inserting the original pixel, that is, the weight of the second original pixel direction is calculated first. (Step S102); using the sexual interpolation between the first original pixel and its adjacent homochromatic pixels to calculate the pixel difference value, "the first set of pixel difference values of the second pixel (step S104). Pixels Each #近28 1282519 first-interpolated pixel of the original pixel (step to get interpolated at step S106); continue to calculate the second set of neighbors of the second original pixel (step measurement), and calculate the second original pixel two image = Weight: , the prime difference is similar to step S1H)), the last generation of the interpolation formula of the present invention # is inserted into the second interpolation pixel of the second original pixel (step S112). When calculating the above-mentioned first-original pixel After inserting the pixels into the second original pixel, the result is obtained. Interpolating the pixels of the three original pixels. Initially, calculating the weight of the first set of adjacent pixel directions of the third original pixel (step S113); the interpolated pixel difference between the image start pixel and the second original pixel in the Liaocai direction a first set of pixel difference values of adjacent pixels of the second original pixel, and substituting the interpolation formula of the present invention to obtain a first interpolated pixel interpolated to the pixel (step Si 17); Then calculating a second set of neighboring pixel weights of the third original pixel (step SU9), and then calculating a second set of pixel difference values of each adjacent pixel of the third original pixel (step S121), and finally, interpolating at the beginning The second interpolated pixel of the pixel (step S123). ', however, the above steps S103 and S104 are evaluated by bilinear interpolation, and each pixel-like color difference value domain will cause an error of insertion, after - In step S125, 'for the time using the bilinear interpolation method, the partial correction of the difference 'recalculates the pixel difference in #, such as K34, ^, K, 5, etc.. Process, with the first element of different colors, the second original pixel (four) three original Based on the pixel, each neighboring element i 29 2982519 = the weight value in each direction 'and calculate the pixel difference between the two, and the interpolated formula of each original pixel is obtained by the interpolation formula of the present invention. As described above, the present invention provides a color interpolation method with directional weights, which improves the edge blur generated by the conventional interpolation method, and effectively interpolates the image edge into a real-unique invention article, and With industrial use: novelty and progressiveness, fully in line with the requirements for invention patent applications, conversion, and requests, please check and grant the patent in this case to protect the inventor's rights.

The above is only a feasible embodiment of the vehicle of the present invention, and therefore, the scope of the patent is not limited to the scope of the patent. Therefore, the equivalent structural changes in the specification and the illustration of the present invention are equally included in Within the scope of the present invention, Chen Ming is incorporated. [Simple diagram of the diagram] The first diagram is not a schematic diagram of a conventional color filter array; the second diagram is a schematic diagram of the edges of the conventional technology image; the second diagram is the image edge color filter adopted by the conventional technique. Schematic diagram of the chip array; the second A picture is a schematic diagram of the edge of the conventional technology image; the second B picture is a schematic diagram of the image edge color filter array adopted in the conventional technology; the fourth A picture is not a schematic diagram of each adjacent pixel embodiment; 4B to EE are diagrams showing the relationship between pixels to be interpolated and their neighboring pixels; FIG. 5 is a schematic diagram of an embodiment of adjacent pixels not being beveled edges; 30 1282519 - Figure 6 is shown in the calculation of the present invention The color filter array with pixel values is not intended. The seventh figure shows a color filter array for calculating the interpolated pixel values according to the present invention. The eighth figure shows the method for calculating the direction weights by using the color difference values. The ninth figure shows a flow chart of an embodiment of the interpolation method of the present invention; the tenth figure shows a flow chart of an embodiment of the interpolation method of the present invention; and the eleventh figure shows the implementation of the interpolation method of the present invention; Example flow chart. [Description of main component symbols] First region 21 Second region 22 Third region 33 Fourth region 34 Left region 41 Right region 42 Upper region 51 Lower region 52 First oblique direction 81 Second oblique direction 82 Third Bevel direction 83 fourth bevel direction 84 31

Claims (1)

1282519 •, the scope of application for patents: 1· - a color interpolation method with directional weights, the method steps include: taking a plurality of original pixels on the couch; generating a digital image of the color filter and the light film array; Calculating a weight of a first set of adjacent pixels of the original pixel; calculating a first set of bilinear interpolation values of the original pixel; calculating a first of a plurality of adjacent pixels of the original pixel and the first set of bilinear interpolation values a pixel difference value; a f-interpolated pixel of the original pixel; a weight of the second set of adjacent pixel directions of the original pixel; a second peak difference of the original pixel; A second interpolated pixel interpolated into the original pixel is derived. 2. f applies for the directional weight value described in item 1 of the patent scope. The weight of the first group of adjacent pixels and the second group of the original pixel is the vertical of the original pixel or Calculated by a plurality of adjacent pixels in the horizontal direction. 3. In the color of the directional weight as described in Item 1 of the patent circumstance (4) = where the color is calculated in the first set of bilinearity of the original pixel; Interpolation, "^, the range is evaluated by the - bilinear interpolation method. :::Special:! The color internal phase with directional weights as described in item 3_"===插法系Calculating the starting edge of the yarn and its 32 1282519 5. The color interpolation method with directional weight as described in claim 1 wherein the first interpolated pixel is weighted by the direction of the first set of adjacent pixels The first set of pixel differences is substituted into the following equation: Yi, i=Xi, j+(Wi-i, jX Ki-i, j + Wi, j-iX Ki, ji + Wi+i, jX Ki+i, j + Wi, j+ix Ki, Chuan) / (Wi-i, J + Wi,] - i + WhuJ + Wi, Sichuan) where i and j are the plurality of original pixels in a color filter array (CFA) The position in the center is the original pixel, and Yi, j is the first interpolated pixel in which the original pixel is interpolated. 6. The color interpolation method with directional weight according to claim 1, wherein the second interpolated pixel is a weight of the second group of adjacent pixels and the second group of pixels Substituting the following equation: Zi, j = Xi, j-(Wi-l, j-lX Ki-1, j-1 + Wi + 1, j-lX Ki + 1, j-1 + Wi + 1, j + lX Ki + 1, j + 1 + Wi-1, j + lX Ki-1, j + l)/(Wi-l, j-1 + Wi + 1, j-1 + Wi + 1, j + 1 + Wi-1, j + l) where i and j are the positions of the plurality of original pixels in a color filter array (CFA), Xu is the original pixel, and Zu is the second of the original pixels interpolated Interpolate pixels. 7. For the color interpolation method with directional weight as described in item 1 of the patent application, the error calculated by the bilinear interpolation method is corrected by recalculating the first group of pixel differences. 8. A color interpolation method with directional weights, the method steps comprising: capturing a plurality of original pixels; generating a digital image of a color filter array; 33 1282519 counting the first group of the original pixels Calculating a weight of a neighboring pixel direction, calculating a first set of bilinear interpolation values of the original pixel, and calculating a first set of pixel difference values between the plurality of neighboring pixels of the original pixel and the first set of bilinear interpolation values; Deriving a first interpolated pixel of the original pixel; calculating a plurality of different values of the original pixel and the adjacent original pixel; ', the weight of the second set of adjacent pixel directions of the counting pixel is based on the color difference value Calculating; calculating a second set of pixel difference values of the original pixel; and deriving a second interpolated pixel of the original pixel. 9. The chromaticity value of the directionality value in the eighth item of the C range is as follows: the first group of adjacent pixels of the original pixel and the weight of the second group = pixel direction are perpendicular to the original pixel or Calculated by a plurality of adjacent pixels in the horizontal direction. Color interpolation with directional weights as described in item 8 = 2 caps The first set of bilinears in the original pixel, the color 帛^ difference value domain performs image interpolation, η. It is evaluated by a bilinear interpolation method. Insertion: The adjacent color-matching method in the color of the directional weights described in the item calculates the in-color direction of the original pixel and its directional weight as described in item 12.=. The first interpolated pixel is obtained from the (four) (four) first-peak difference field of the first group of neighboring pixels. 34 1282519 · Out: Yi, j = Xi, j+(Wi-l, jX Ki-1 , j + Wi, j-lX Ki, j-1 + Wi + 1, jX Ki + 1, j + Wi, j + 1 X Ki, j + l)/(Wi-l, j + Wi, j-1 + Wi + 1, j + Wi, j + l) where i and j are the positions of the plurality of original pixels in a color filter array • (CFA), Xu is the original pixel, and Yi, j is interpolated The first interpolated pixel of the original pixel. 13. The color interpolation method with directional weight according to claim 8, wherein the second interpolated pixel is different from the second group of pixels by the weight of the second group of adjacent pixels. The value is substituted into the following equation: Zi, j^Xi, j-(Wi-l, j-lX Ki-1, j-1 + Wi + 1, j-lX Ki + 1, j-1 + Wi + 1, j + lX Ki + 1, j + l + Wi-1, j + lX Ki-1, j + l)/ (Wi-1, j-1 + Wi + 1, j-1 + Wi + 1, j + l + Wi-1, j + l) where i and j are the positions of the plurality of original pixels in a color filter array (CFA), 乂 "for the original pixel, Zu is the interpolation of the original pixel The second interpolated pixel. 14. The intra-color φ interpolation method with directional weight as described in claim 8 of the patent application, further recalculating the first set of pixel differences for correction using the bilinear interpolation The error calculated by the method. 15. The color interpolation method with directional weight as described in claim 8 recalculates the first group of pixel differences by calculating the color difference value of the original pixel and the adjacent pixel. Value to correct the color of the fall. 16. - Directions a color interpolation method of weights, the method steps comprising: capturing a plurality of original pixels; generating a digital image of a color filter array; 35 1282519 calculating a weight of a first set of adjacent pixel directions of a first original pixel Calculating a first set of pixel difference values of each adjacent pixel of the first original pixel; obtaining a first interpolated pixel interpolated in one of the first original pixels; calculating a second set of adjacent pixel directions of the first original pixel a second set of pixel difference values of each adjacent pixel of the first original pixel; a second interpolated pixel interpolated to one of the first original pixels; and a first set of adjacent pixels of a second original pixel Calculating a weight of the direction; calculating a first set of pixel difference values of each adjacent pixel of the second original pixel; obtaining a first interpolated pixel of the second original pixel; and discriminating the second set of adjacent pixels of the second original pixel Calculating a weight of the direction; calculating a second set of pixel difference values of each adjacent pixel of the second original pixel; and applying a second interpolated pixel interpolated to one of the second original pixels; a weight of the first group of adjacent pixels; a first set of pixel differences of adjacent pixels of the second pixel; a first interpolated pixel interpolated to one of the third original pixels; a weight of a second set of adjacent pixel directions of the original pixel; a second set of pixel difference values of each adjacent pixel of the third original pixel; and a second interpolated pixel interpolated to one of the third original pixels π· The color interpolation method with directional weight as described in claim 16 wherein the weight of the first set of adjacent pixels of the first original pixel and the direction of the first set of adjacent pixels is Calculated by a plurality of adjacent pixels in the vertical or horizontal direction of the original pixel. 18. Within the color of the directional weight as described in claim 16 of the patent scope 36 1282519 near C, the first group of neighboring pixels of the original pixel and the weight of the near-original direction are borrowed by the first-original pixel and the Neighbor, ,: The multiple color differences of pixels are worthy of being worth. The color of the directional weight value of the two sets of adjacent ΐιϊ second original pixels in the color of the insert i is specified by the first pixel of the second original pixel and the weight in the horizontal direction is perpendicular to the original pixel or The last few neighboring pixels on the 9 gate are calculated. The value in the horizontal direction of the color of the directional weight described in item 16 is calculated from the vertical pixels of the original pixel or a hard number of adjacent pixels on the gate. • Insert 2 special ==: the step of the color interpolation value of the directional weight, the first set of bilinearity within the first linearity of the color difference 差异 ~ difference value domain to implement image interpolation, 22. The domain is evaluated by a linear interpolation method. The adjacent color in the color of the directional weight described in the interpolation term is: = interpolation method calculates the color of the original ^ 23:5 special: = 16 items with directional weights - group proximity The first-interpolated pixel of the if-square pixel is substituted by the first-order equation ΐΐί and the first-group pixel difference into Xy+dijX Ki-i j + Wi 】] χ κ Ax Kl+1&gt ; J + Wl'J+1 where 1 and j are the positions of the plurality of original pixels in a color filter array 37 1282519 (CFA), and Xi'j is the first original pixel, the second original, and the pixel Or the third original pixel, Yu is the first interpolated pixel interpolating the original pixel. 24. The color interpolation method of directional weight according to claim 16, wherein the second interpolated pixel of each original pixel is weighted by the second set of adjacent pixel directions and the second The group pixel difference is substituted into the following equation: Zi^^j-(Wi-i, MX Ki-l, M + Wi + l, MX Ki+l, M + Wi + i, j + lX Ki + l, J + l + Wi-l, j + lX Ki-1, j + l)/(Wi-l, j-1 + Win, j-1 + WiH, j + l + Wi-1, j + l) where i and j are the plurality of original pixels in one The position in the color filter array (CFA), Xu is the first original pixel, the second original pixel or the third original pixel, and Zi, j is the second interpolated pixel interpolating the original pixel. X 25· The color interpolation method with directional weight as described in claim 16 of the patent application is performed by recalculating the plurality of pixel differences to correct the error calculated by the bilinear interpolation method. 26. The color interpolation method of directional weight according to claim 16, wherein the first original pixel and the first interpolated pixel of the second original pixel are bilinearly interpolated. To evaluate the inter-pixel = difference value field and correct for the partial use of the bilinear interpolation method. With the morphological weight insertion method described in item 16 of the patent application scope, the difference between the original pixel and the adjacent pixel is recalculated to correct the pixel difference value. The color of the fall. Shape difference 38