CN101917629B - Green component and color difference space-based Bayer format color interpolation method - Google Patents

Abstract

The invention relates to a green component and color difference space-based Bayer format color interpolation method. Two main errors which are jaggies and pseudo colors respectively exist in the conventional color interpolation methods. In the method, in a full-color image recovery process, color recovery is realized by adopting a combined way of interpolating green components, red components and blue components in steps; and the method comprises the reconstruction of the full-resolution green components G, the calculation of pixel point color differences R-G or B-G comprising the red or blue components, the reconstruction of full-resolution color difference images R-G and B-G, and the recovery of the red components R and the blue components B by the addition of the reconstructed full-resolution green components and the reconstructed color difference images. The method of the invention has the advantages of effectively solving the problem that the images have blurry edges and unsharp detail textures, reducing the color distortion of the images, preventing color mutation and improving color smoothness.

Description

A kind of Bayer form color interpolation method based on green component and aberration space

Technical field

The invention belongs to the digital image processing techniques field, relate to a kind of Bayer form color interpolation method based on green component and aberration space, specifically is the method for the compound mode realization color restoration of a kind of green component and red blue component substep interpolation.

Background technology

The continuous maturation of Along with computer technology is digitized into the trend into social development, is the attention that the digital imaging technology of representative has caused academia and industrial circle with the digital camera.In in the past 20 years, during CCD and cmos digital camera are applied to daily life more and more and produce.Consider cost and volume, most of digital camera generally adopts single CCD or CMOS as the transducer that receives image.In order accurately intactly to describe a color pixel cell, need three kinds of color components of red, green, blue at least, therefore can on the digital camera imageing sensor, cover the color filter array.Because Bayer form color filter array has good color signal sensitiveness and color restoration characteristic, therefore be applied in widely on the various imageing sensors.Each sensitive spot in the Bayer form color filter array only allows a kind of color component to pass through, so each pixel of image all has only a kind of color gray value, and is as shown in Figure 1.Therefore, for the coloured image that regains one's integrity, need calculate other two kinds of disappearance color components through the color around each sampled point of color interpolation techniques make use.

In the digital camera system, most of color filter arrays have only a kind of color component that higher sample frequency is arranged at present, and it is also more to compare other two kinds of color component samplings.In Bayer form color filter array, green component (G) is account for the total sampled point of image half the, is the twice of red component (R) or blue component (B), therefore comprises more images information, and human eye is red and blue responsive to green.So the green component of present general first each pixel of reconstructed image of most color interpolation algorithm, and then rebuild other two kinds of color components.More typical color interpolation algorithm has following three kinds at present:

Bilinearity (Bilinear) interpolation algorithm: early stage owing to receive the restriction of hardware condition, be more prone to adopt that algorithm is simple, the interpolation algorithm of fast operation, bilinear interpolation algorithm a kind of algorithm of being used widely that comes to this.This algorithm utilizes the to be asked color component of the mean value of color component of the same race in the neighbor sampled point as current pixel.

Edge guiding (Edge Adaptive) interpolation algorithm: in the Interpolation Process, at first on the comparison level direction with vertical direction on gradient magnitude, the pixel of getting during interpolation on the less direction of gradient calculates a color component value current to be asked as estimation point.In order to reduce algorithm complex as far as possible, at first utilize the green component of this each pixel of algorithm reconstructed image, utilize bilinearity then and the compensation correction of the green component rebuild recovers redness and blue component.

Adaptability color layers (Adapted Color Plane) interpolation algorithm: this algorithm leads on the edge of to improve on the interpolation algorithm and obtains; Different is that this method is when considering the image border; Confirm through green component G and red R or blue B component, and all adopt this algorithm to carry out interpolation to three kinds of components of red, green, blue and recover.Likewise, rebuild earlier the green component of full resolution, utilize the reconstruction of the red and blue component of the green component correction of having rebuild again.

Bilinear interpolation algorithm the earliest is not owing to there is the edge of differentiate between images in computational process; Just simply average interpolation through contiguous pixel; The image restored effect is unsatisfactory; Edge distortion is apparent in view, and image detail and texture performance are not accurate enough, and have produced serious color distortion.In order to overcome problems such as soft edge, be greatly improved than the bilinear interpolation algorithm based on the edge guiding interpolation algorithm texture of gradient and the performance at edge.And it is more accurate to the location of image border to improve the adaptability color layers interpolation algorithm that obtains on the interpolation algorithm that leads on the edge of, has further reduced the fuzzy and cross-color of edge of image, and the image restored effect is more satisfactory.Though this algorithm has solved the soft edge problem, do not consider the color high correlation of image, can not prevent the color sudden change of image effectively, effect is relatively poor aspect the color smoothness.

Summary of the invention

The object of the present invention is to provide a kind of Bayer form color interpolation method based on green component and aberration space, it is a kind of novel Bayer format-pattern color interpolation method that makes full use of image space correlation and color correlation.This method is recovered the green component of disappearance earlier, and the correction-compensation that utilizes the green component that has recovered again is to rebuild red blue component, simultaneously in the process of rebuilding red blue component; Introduced image aberration rule; Be specifically designed to the false colour that produces in the inhibition Interpolation Process, prevent the color saltus step, strengthen the color smoothness; Reduce the fog-level at edge, improve the performance accuracy of image texture and details.

Bayer format-pattern color interpolation method provided by the invention adopts the compound mode of green component and red blue component substep interpolation to realize color restoration in the process of rebuilding full-color image.At first utilize the strict differentiate between images of the spatial coherence edge of image, rebuild the full resolution green component; Utilize the color correlation of image to carry out the interpolation calculation in aberration (R-G and B-G) space then, rebuild the red blue component of full resolution through the correction-compensation of the green component that recovered again.Than interpolation algorithms such as Traditional use bilinearity, edge guiding and adaptability color layers, the full-color image effect that recovery obtains is even more ideal, and the Y-PSNR of color of image signal is also higher, also has the low advantage of algorithm computational complexity simultaneously.

The inventive method comprises reconstruction and four contents of reconstruction of full resolution red component R and blue component B of reconstruction, the calculating at pixel aberration R-G that comprises redness or blue component or B-G, full resolution color difference image R-G and the B-G of full resolution green component G.

Step 1. is rebuild full resolution green component G

In Bayer form color filter array, the green component sampling number accounts for the half the of the total sampling number of image, is the twice of red component or blue component sampling number; Therefore comprise more images information; And human eye is red and blue responsive to green, to brightness than color sensitivity, and in the color space conversion from RGB to YUV; Near 60%, brightness Y calculates green component to the contribution of brightness Y:

Y＝0.299R+0.587G+0.114B. (1)

Therefore the full resolution green component be reconstituted in the full-color image recovery process most importantly, its concrete grammar is:

At the pixel sampling point place that comprises red component R, take all factors into consideration the green component G and the red component R of current point neighborhood, calculated level direction gradient H from horizontal direction and vertical direction _GrWith vertical gradient V _Gr, select the edge direction of the less direction of gradient as the current pixel sampled point.

1) works as H _Gr＞V _GrThe time, then vertical direction selects two adjacent on current point vertical direction green component G and two red component R and current red component R to calculate the green component G that current point lacks as the edge direction of current point;

2) work as H _Gr＜V _GrThe time, then horizontal direction selects two adjacent on current point horizontal direction green component G and two red component R and current red component R to calculate the green component G that current point lacks as the edge direction of current point;

3) work as H _Gr=V _GrThe time, then select four adjacent green component G of current point and four red component R and current red component R to calculate the green component G that current point lacks on every side.

At the pixel sampling point place that comprises blue component B, take all factors into consideration the green component G and the blue component B of current point neighborhood, calculated level direction gradient H from horizontal direction and vertical direction _GbWith vertical gradient V _Gb, select the edge direction of the less direction of gradient as current point.

1) works as H _Gb＞V _GbThe time, then vertical direction selects two adjacent on current point vertical direction green component G and two blue component B and current blue component B to calculate the green component G that current point lacks as the edge direction of current point;

2) work as H _Gb＜V _GbThe time, then horizontal direction selects two adjacent on current point horizontal direction green component G and two blue component B and current blue component B to calculate the green component G that current point lacks as the edge direction of current point;

3) work as H _Gb=V _GbThe time, then select four adjacent green component G of current point and four blue component B and current blue component B to calculate the green component G that current point lacks on every side.

Step 2. is calculated the pixel aberration R-G that comprises red component and is calculated the pixel aberration B-G that comprises blue component.

In different Color plane, have high correlation between each chrominance component, just color ratio rule and aberration rule, the color-values of pixel is more approaching in a little neighborhood.The present invention utilizes the aberration rule to carry out the recovery of red component R and blue component B, at first will calculate the color difference image R-G and the B-G of Bayer form, its concrete grammar:

Comprising the pixel sampling point of red component R, on the basis of step 1, the green component G that current point includes original red color component R and rebuild, calculated difference R-G obtains the red green difference image of Bayer form.

Comprising the pixel sampling point of blue component B, on the basis of step 1, the green component G that current point includes original blue component B and rebuild, calculated difference B-G obtains the blue-green difference image of Bayer form.

Step 3. is rebuild full resolution color difference image R-G and color difference image B-G

The red green difference R-G and the bluish-green aberration B-G image of Bayer form in step 2, have been obtained.Red green almost comprises red component R sample point in original Bayer formatted data, blue-green almost comprises blue component B sample point in original Bayer formatted data.Through step 1 obtain rebuild full resolution green component G and current point image border confirm to rebuild respectively full resolution color difference image R-G and B-G, its detailed process:

(1), rebuild the concrete grammar of full resolution color difference image R-G:

Comprising aberration B-G but the sample point of disappearance aberration R-G; Adjacent on every side has four red greens not good enough; Lay respectively at upper left, left following, bottom right, upper right; Following two miter angle tilted directions are taken all factors into consideration the green component G and the red green difference R-G of current point neighborhood from a upper left bottom right and a upper right left side, calculate upper left lower right respectively to Grad H _R-gWith upper right lower left to Grad V _R-g, select the edge direction of the less direction of Grad as current point.

1) works as H _R-g＞V _R-gThe time; As the edge direction of current point, the green component G that two red green difference R-G that select that the upper right lower left of current point makes progress adjacent and two green component G that rebuild and current point have been rebuild calculates the red green difference R-G that current point lacks under the then upper right left side;

2) work as H _R-g＜V _R-gThe time; Then upper left bottom right is as the edge direction of current point, and the green component G that two red green difference R-G that select that the upper left lower right of current point makes progress adjacent and two green component G that rebuild and current point have been rebuild calculates the red green difference R-G that current point lacks;

3) work as H _R-g=V _R-gThe time, the green component G that then selects four adjacent red green difference R-G of two miter angle tilted directions of current point and four green component G that rebuild and current point to rebuild calculates the red green difference R-G that current point lacks.

At the pixel sampling point place that only comprises green component G; Adjacent on every side has two red greens not good enough; Employing is got mean deviation to two adjacent red green difference R-G and is revised with not good enough green component and the current point green component G that locates of adjacent red green, obtains the red green difference R-G that current point lacks.

(2), rebuild the concrete grammar of full resolution color difference image B-G:

Comprising aberration R-G but the sample point of disappearance aberration B-G; Adjacent on every side has four red greens not good enough; Lay respectively at upper left, left following, bottom right, upper right; Following two miter angle tilted directions are taken all factors into consideration the green component G and the bluish-green aberration B-G of current point neighborhood from a upper left bottom right and a upper right left side, calculate upper left lower right respectively to Grad H _B-gWith upper right lower left to Grad V _B-g, select the edge direction of the less direction of Grad as current point.

1) works as H _B-g＞V _B-gThe time; As the edge direction of current point, the green component G that two red green difference R-G that select that the upper right lower left of current point makes progress adjacent and two green component G that rebuild and current point have been rebuild calculates the red green difference R-G that current point lacks under the then upper right left side;

2) work as H _B-g＜V _B-gThe time; Then upper left bottom right is as the edge direction of current point, and the green component G that two red green difference R-G that select that the upper left lower right of current point makes progress adjacent and two green component G that rebuild and current point have been rebuild calculates the red green difference R-G that current point lacks;

3) work as H _B-g=V _B-gThe time, the green component G that then selects four adjacent red green difference R-G of two miter angle tilted directions of current point and four green component G that rebuild and current point to rebuild calculates the red green difference R-G that current point lacks.

Only comprising the pixel sampling point of green component G; Adjacent on every side has two blue-greens not good enough; Employing is got mean deviation to two adjacent bluish-green aberration B-G and is revised with not good enough green component and the current point green component G that locates of adjacent blue-green, obtains the bluish-green aberration B-G that current point lacks.

Step 4. is rebuild full resolution red component R and blue component B

Obtained full resolution color difference image R-G and B-G in step 3, promptly each pixel sampling point of image all includes red green difference R-G and blue-green difference B-G.Obtained full resolution green component G in step 1, promptly each pixel sampling point of image all includes green component G.The concrete grammar of rebuilding full resolution red component R is:

With red green difference R-G in each pixel sampling point of image and the green component G addition in the current sampling point, calculate all sampled points of image and promptly obtained full resolution red component R, promptly each pixel sampling point of image all includes red component R.

The concrete grammar of rebuilding full resolution blue component B is:

With blue-green difference B-G in each pixel sampling point of image and the green component G addition in the current sampling point, calculate all sampled points of image and promptly obtained full resolution blue component B, promptly each pixel sampling point of image all includes blue component B.

The inventive method effectively reduces in the color interpolation process and to produce edge blurry and color distortion, has solved that the image effect that bilinear interpolation method brought is undesirable, edge blurry, image texture and problem such as the details performance is inaccurate, color distortion is serious.Edge guiding commonly used is more accurate with adaptability color layers interpolation method imaging effect is better, the edge is judged, the color smoothness is higher and each color component Y-PSNR of image restored is higher, and computing cost is little.

Description of drawings

Fig. 1 is a Bayer form color filter array sketch map;

Fig. 2 is a flow chart of the present invention;

Fig. 3 is the reconstruction strategy sketch map of red, green, blue three primary colors of the present invention;

Fig. 4 is 6 * 6 original reference pixel sampling point sketch mapes of Bayer form;

6 * 6 green component G sketch mapes of Fig. 5 for having rebuild;

Fig. 6 is 6 * 6 corresponding sampled point aberration sketch map as a result of Bayer form;

6 * 6 red green difference R-G sketch mapes of Fig. 7 for having rebuild;

6 * 6 bluish-green aberration B-G sketch mapes of Fig. 8 for having rebuild;

6 * 6 red component R sketch mapes of Fig. 9 for having rebuild;

6 * 6 blue component B sketch mapes of Figure 10 for having rebuild.

Embodiment

Below in conjunction with accompanying drawing the present invention is described further.

The present invention is applicable to the color interpolation in any Bayer format digital camera imaging system, and Fig. 1 has showed the arrangement mode figure of each color component of Bayer form hypograph.

In the elaborate process of embodiment, symbol R, G, B be red component, green component, the blue component of presentation video respectively, symbol R _Ij, G _Ij, B _IjRespectively red component, green component, the blue component of the capable and j row pixel of presentation video i, symbol R-G, B-G represent redness and green difference, blueness and green difference, symbol R respectively _Ij-G _Ij, B _Ij-G _IjRespectively redness and green difference, blueness and the green difference of the capable and j row pixel of presentation video i, and image is listed as since the 0th row and the 0th.

Fig. 2 and Fig. 3 represent the reconstruction strategy sketch map of schematic flow sheet of the present invention and red, green, blue three primary colors respectively; Recover to obtain full-color image from the Bayer original image; The first step is to rebuild full resolution green component G; Second step was to calculate aberration R-G or B-G at the pixel that comprises redness or blue component; The 3rd step was to rebuild full resolution color difference image R-G and B-G, was that the green component G that has rebuild is recovered red component R and blue component B with color difference image R-G that has rebuild and B-G respectively mutually at last.

The reconstruction of step 1. full resolution green component G

(1), rebuild the green component G of disappearance, take all factors into consideration the green component G and the red component R of current point neighborhood, calculated level direction gradient H from horizontal direction and vertical direction at the pixel sampling point that comprises red component R _GrWith vertical gradient V _GrCurrent point is R _Ij, the gradient H that then image i is capable and j is listed as _GrijAnd V _GrijFor:

H _grij＝|-R _i(j-2)+2R _ij-R _i(j+2)|+|G _i(j-1)-G _i(j+1)|. (2)

V _grij＝|-R _(i-2)j+2R _ij-R _(i+2)j|+|G _(i-1)j-G _(i+1)j|. (3)

Above two Grad are selected the edge direction of less direction as current point.

1) works as H _Grij＜V _GrijThe time, promptly horizontal direction edge difference is less, and then horizontal direction is as the edge direction of current point, choice point R _{I (j-2)}, R _Ij, R _{I (j+2)}, g _{I (j-1)}And G _{I (j+1)}Calculate current point R _IjThe green component G that is lacked _Ij, computational methods are following:

$G_{\mathrm{ij}}=\frac{G_{i(j-1)}+G_{i(j+1)}}{2}+\frac{(-R_{i(j-2)}+2R_{\mathrm{ij}}-R_{i(j+2)})}{4}.---\left(4\right)$

2) work as H _Grij＞V _GrijThe time, promptly vertical direction edge difference is less, and then vertical direction is as the edge direction of current point, choice point R _{(i-2) j}, R _Ij, R _{(i+2) j}, G _{(i-1) j}And G _{(i+1) j}Calculate current point R _IjThe green component G that is lacked _Ij, computational methods are following:

$G_{\mathrm{ij}}=\frac{G_{(i-1)j}+G_{(i+1)\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HSA00000226843300021.png"\; state="\{\{state\}\}">j</figure-callout>}}}{2}+\frac{(-R_{(i-2)j}+2R_{\mathrm{ij}}-R_{(i+2)j})}{4}.---\left(5\right)$

3) work as H _Grij=V _GrijThe time, then select R _{(i-2) j}, R _{I (j-2)}, R _Ij, R _{(i+2) j}, R _{I (j+2)}, G _{(i-1) j}, G _{J (j-1)}, G _{(i+1) j}And G _{I (j+1)}Calculate current point R _IjThe green component G that is lacked _Ij, computational methods are following:

$G_{\mathrm{ij}}=\frac{G_{(i-1)j}+G_{i(j-1)}+G_{(i+1)j}+G_{i(j+1)}}{4}$

$+\frac{(-R_{(i-2)j}-R_{i(j-2)}+4R_{\mathrm{ij}}-R_{(i+2)j}-R_{i(j+2)})}{8}.---\left(6\right)$

(2), rebuild the green component G of disappearance, take all factors into consideration the green component G and the blue component B of current point neighborhood, calculated level direction gradient H from horizontal direction and vertical direction at the pixel sampling point that comprises blue component B _GbWith vertical gradient V _GbCurrent point is B _Ij, the gradient H that then image i is capable and j is listed as _GbijAnd V _GbijFor:

H _gbij＝|-B _i(j-2)+2B _ij-B _i(j+2)|+|G _i(j-1)-G _i(j+1)|. (7)

V _gbij＝|-B _(i-2)j+2B _ij-B _(i+2)j|+|G _(i-1)j-G _(i+1)j|. (8)

Above two Grad are selected the edge direction of less direction as current point.

1) works as H _Gbij＜V _GbijThe time, promptly horizontal direction edge difference is less, and then horizontal direction is as the edge direction of current point, choice point B _{I (j-2)}, B _Ij, B _{I (j+2)}, G _{I (j-1)}And G _{I (j+1)}Calculate current point B _IjThe green component G that is lacked _Ij, computational methods are following:

$G_{\mathrm{ij}}=\frac{G_{i(j-1)}+G_{i(j+1)}}{2}+\frac{(-B_{i(j-2)}+2B_{\mathrm{ij}}-B_{i(j+2)})}{4}.---\left(9\right)$

2) work as H _Gbij＞V _GbijThe time, promptly vertical direction edge difference is less, and then vertical direction is as the edge direction of current point, choice point B _{(i-2) j}, B _Ij, B _{(i+2) j}, G _{(i-1) j}And G _{(i+1) j}Calculate current point B _IjThe green component G that is lacked _Ij, computational methods are following:

$G_{\mathrm{ij}}=\frac{G_{(i-1)j}+G_{(i+1)\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="HSA00000226843300021.png"\; state="\{\{state\}\}">j</figure-callout>}}}{2}+\frac{(-B_{(i-2)j}+2B_{\mathrm{ij}}-B_{(i+2)j})}{4}.---\left(10\right)$

3) work as H _Gbij=V _GbijThe time, then select B _{(i-2) j}, B _{I (j-2)}, B _Ij, B _{(i+2) j}, B _{I (j+2)}, G _{(i-1) j}, G _{I (j-1)}, G _{(i+1) j}And G _{I (j+1)}Calculate current point B _IjThe green component G that is lacked _Ij, computational methods are following:

$G_{\mathrm{ij}}=\frac{G_{(i-1)j}+G_{i(j-1)}+G_{(i+1)j}+G_{i(j+1)}}{4}$

$+\frac{(-B_{(i-2)j}-B_{i(j-2)}+4B_{\mathrm{ij}}-B_{(i+2)j}-B_{i(j+2)})}{8}.---\left(11\right)$

Fig. 4 has provided 6 * 6 original reference pixel sampling point sketch mapes of a Bayer form, further the reconstruction of full resolution green component G is described and describes below in conjunction with the concrete arrangement mode of data of Fig. 4, so that understand method of the present invention.

(1), rebuild the green component G of disappearance, take all factors into consideration the green component G and the red component R of current point neighborhood, calculated level direction gradient H from horizontal direction and vertical direction at the pixel sampling point that comprises red component R _GrWith vertical gradient V _Gr

With current pixel point R ₂₂Be example, then the gradient H of image the 2nd row and the 2nd row _Gr22And V _Hr22For:

H _gr22＝|-R ₂₀+2R ₂₂-R ₂₄|+|G ₂₁-G ₂₃|. (12)

V _gr22＝|-R ₀₂+2R ₂₂-R ₄₂|+|G ₁₂-G ₃₂|. (13)

Above two Grad are selected the edge direction of less direction as current point.

1) works as H _Gr22＜V _Gr22The time, promptly horizontal direction edge difference is less, and then horizontal direction is as the edge direction of current point, choice point R ₂₀, R ₂₂, R ₂₄, G ₂₁And G ₂₃Calculate current point R ₂₂The green component G that is lacked ₂₂, computational methods are following:

$G_{22}=\frac{G_{21}+G_{23}}{2}+\frac{(-R_{20}+2R_{22}-R_{23})}{4}.---\left(14\right)$

2) work as H _Gr22＞V _Gr22The time, promptly vertical direction edge difference is less, and then vertical direction is as the edge direction of current point, choice point R ₀₂, R ₂₂, R ₄₂, G ₁₂And G ₃₂Calculate current point R ₂₂The green component G that is lacked ₂₂, computational methods are following:

$G_{22}=\frac{G_{12}+G_{32}}{2}+\frac{(-R_{02}+2R_{22}-R_{42})}{4}.---\left(15\right)$

3) work as H _Gr22=V _Gr22The time, then select R ₀₂, R ₂₀, R ₂₂, R ₄₂, R ₂₄, G ₁₂, G ₂₁, G ₃₂And G ₂₃Calculate current point R ₂₂The green component G that is lacked ₂₂, computational methods are following:

$G_{22}=\frac{G_{12}+G_{21}+G_{32}+G_{23}}{4}+\frac{(-R_{02}-R_{20}+4R_{22}-R_{42}-R_{24})}{8}.---\left(16\right)$

With special pixel point R ₀₂Be example, then the gradient H of image the 0th row and the 2nd row _Gr02And V _Gr02For:

H _gr02＝|-R ₀₀+2R ₀₂-R ₀₄|+|G ₀₁-G ₀₃|. (17)

V _gr02＝|R ₀₂-R ₂₂|+|G ₁₂|. (18)

Above two Grad are selected the edge direction of less direction as current point.

1) works as H _Gr02＜V _Gr02The time, promptly horizontal direction edge difference is less, and then horizontal direction is as the edge direction of current point, choice point R ₀₀, R ₀₂, R ₀₄, G ₀₁And G ₀₃Calculate current point R ₀₂The green component G that is lacked ₀₂, computational methods are following:

$G_{02}=\frac{G_{01}+G_{03}}{2}+\frac{(-R_{00}+2R_{02}-R_{04})}{4}.---\left(19\right)$

2) work as H _Gr02＞V _Gr02The time, promptly vertical direction edge difference is less, and then vertical direction is as the edge direction of current point, choice point R ₀₂, R ₂₂, G ₁₂Calculate current point R ₀₂The green component G that is lacked ₀₂, computational methods are following:

$G_{02}=G_{12}+\frac{(R_{02}-R_{22})}{2}.---\left(20\right)$

3) work as H _Gr02=V _Gr02The time, then select R ₀₀, R ₀₂, R ₂₂, R ₀₄, G ₁₂, G ₀₁And G ₀₃Calculate current point R ₀₂The green component G that is lacked ₀₂, computational methods are following:

$G_{02}=\frac{G_{12}+G_{01}+G_{03}}{3}+\frac{(-R_{00}+3R_{02}-R_{22}-R_{04})}{6}.---\left(21\right)$

(2), rebuild the green component G of disappearance, take all factors into consideration the green component G and the blue component B of current point neighborhood, calculated level direction gradient H from horizontal direction and vertical direction at the pixel sampling point that comprises blue component B _GbWith vertical gradient V _Gb

With current pixel point B ₃₃Be example, then the gradient H of image the 3rd row and the 3rd row _Gb33And V _Gb33For:

H _gb33＝|-B ₃₁+2B ₃₃-B ₃₅|+|G ₃₂-G ₃₄|. (22)

V _gb33＝|-B ₁₃+2B ₃₃-B ₅₃|+|G ₂₃-G ₄₃|. (23)

Above two Grad are selected the edge direction of less direction as current point.

1) works as G _Gb33＜V _Gb33The time, promptly horizontal direction edge difference is less, and then horizontal direction is as the edge direction of current point, choice point B ₃₁, B ₃₃, B ₃₅, G ₃₂And G ₃₄Calculate current point B ₃₃The green component G that is lacked ₃₃, computational methods are following:

$G_{33}=\frac{G_{32}+G_{34}}{2}+\frac{(-B_{31}+2B_{33}-B_{35})}{4}.---\left(24\right)$

2) work as H _Gb33＞V _Gb33The time, promptly vertical direction edge difference is less, and then vertical direction is as the edge direction of current point, choice point B ₁₃, B ₃₃, B ₅₃, G ₂₃And G ₄₃Calculate current point B ₃₃The green component G that is lacked ₃₃, computational methods are following:

$G_{33}=\frac{G_{23}+G_{43}}{2}+\frac{(-B_{13}+{2B}_{33}-B_{53})}{4}.---\left(25\right)$

3) work as H _Gb33=V _Gb33The time, then select B ₁₃, B ₃₁, B ₃₃, B ₅₃, B ₃₅, G ₃₂, G ₃₄, G ₂₃And G ₄₃Calculate current point B ₃₃The green component G that is lacked ₃₃, computational methods are following:

$G_{33}=\frac{G_{23}+G_{32}+G_{34}+G_{43}}{4}+\frac{(-B_{13}-B_{31}+4B_{33}-B_{35}-B_{53})}{8}.---\left(26\right)$

With special pixel point B ₅₅Be example, then the gradient H of image the 5th row and the 5th row _Gb55And B _Gb55For:

H _gb55＝|B ₅₅-B ₅₃|+|G ₅₄|. (27)

V _gb55＝|B ₅₅-B ₃₅|+|G ₄₅|. (28)

Above two Grad are selected the edge direction of less direction as current point.

1) works as H _Gb55＜V _Gb55The time, promptly horizontal direction edge difference is less, and then horizontal direction is as the edge direction of current point, choice point B ₅₃, B ₅₅, and G ₅₄Calculate current point B ₅₅The green component G that is lacked ₅₅, computational methods are following:

$G_{55}=G_{54}+\frac{(B_{55}-B_{53})}{2}.---\left(29\right)$

2) work as H _Gb55＞V _Gb55The time, promptly vertical direction edge difference is less, and then vertical direction is as the edge direction of current point, choice point B ₃₅, B ₅₅, and G ₄₅Calculate current point B ₅₅The green component G that is lacked ₅₅, computational methods are following:

$G_{55}=G_{45}+\frac{(B_{55}-B_{35})}{2}.---\left(30\right)$

3) work as H _Gb55=V _Gb55The time, then select B ₃₅, B ₅₃, B ₅₅, G ₅₄And G ₄₅Calculate current point B ₅₅The green component G that is lacked ₅₅, computational methods are following:

$G_{55}=\frac{G_{45}+G_{54}}{2}+\frac{(-B_{35}+{2B}_{55}-B_{53})}{4}.---\left(31\right)$

The span of green component G is 0 to 255, therefore when more than the green component G that calculates need overflow compensation deals when exceeding this span, promptly when the result of calculation of G less than 0 the time, the then current green component G that rebuilds is 0; When the result of calculation of G greater than 255 the time, the then current green component G that rebuilds is 255.Fig. 5 has showed 6 * 6 green component G that rebuild that calculate through step 1.

Step 2. is in pixel aberration R-G that comprises redness or blue component or the calculating of B-G

On the basis of step 1, obtained the full resolution green component G that rebuild, promptly each pixel sampling point of image all includes green component G.

(1), comprising the pixel sampling point of red component R, current point is R _Ij, the green component that comprises is G _Ij, then image i red green differences capable and the j row are R _Ij-G _Ij

(2), comprising the pixel sampling point of blue component B, current point is B _Ij, the green component that comprises is G _Ij, then image i bluish-green aberration capable and the j row are B _Ij-G _Ij

Further the calculating of aberration R-G or B-G is described and describes below in conjunction with the concrete arrangement mode of data of Fig. 4 and the green component of rebuilding of Fig. 5, so that understand method of the present invention.

(1), comprising the pixel sampling point of red component R, with the R of Fig. 4 ₂₂G with Fig. 5 ₂₂Be example, then the red green difference of image the 2nd row and the 2nd row is R ₂₂-G ₂₂

(2), comprising the pixel sampling point of blue component B, with the B of Fig. 4 ₅₅G with Fig. 5 ₅₅Be example, then the bluish-green aberration of image the 5th row and the 5th row is R ₅₅-G ₅₅

6 * 6 respective pixel sampled point aberration that Fig. 6 has showed the Bayer form that calculates through step 2 are sketch map as a result.

The reconstruction of step 3. full resolution color difference image R-G and B-G

On the basis of step 2, obtained the red green difference R-G and the bluish-green aberration B-G image of Bayer form.

(1), the reconstruction of full resolution color difference image R-G

A), comprising aberration B-G but the sample point of disappearance aberration R-G; Adjacent on every side has four red greens not good enough; Lay respectively at upper left, left following, bottom right, upper right; Following two miter angle tilted directions are taken all factors into consideration the green component G and the red green difference R-G of current point neighborhood from a upper left bottom right and a upper right left side, calculate upper left lower right respectively to Grad H _R-gWith upper right lower left to Grad V _R-gCurrent point is B _Ij-G _Ij, the aberration gradient H that then image i is capable and j is listed as _{(r-g) ij}And V _{(r-g) ij}For:

H _(r-g)ij＝|(R _(i-1)(j-1)-G _(i-1)(j-1))-(R _(i+1)(j+1)-G _(i+1)(j+1))|

(32)

+|-G _(i-1)(j-1)+2G _ij-G _(i+1)(j+1)|.

V _(r-g)ij＝|(R _(i-1)(j+1)-G _(i-1)(j+1))-(R _(i+1)(j-1)-G _(i+1)(j-1))|

(33)

+|-G _(i-1)(j+1)+2G _ij-G _(i+1)(j-1)|.

Above two Grad are selected the edge direction of less direction as current point.

1) works as H _{(r-g) ij}＜V _{(r-g) ij}The time, then upper left bottom right is as the edge direction of current point, choice point R _{(i-1) (j-1)}-G _{(i-1) (j-1)}, R _{(i+1) (j+1)}-G _{(i+1) (j+1)}, G _{(i-1) (j-1)}, G _IjAnd G _{(i+1) (j+1)}Calculate current point B _Ij-G _IjThe red green difference R that is lacked _Ij-G _Ij, computational methods are following:

$R_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(R_{(i-1)(j-1)}-G_{(i-1)(j-1)})+\left(R_{(i+1)\left(j\right)+1}\right)-G_{(i+1)(j+1)}}{2}$

$+\frac{(-G_{(i-1)(j-1)}+2G_{\mathrm{ij}}-G_{(i+1)(j+1)})}{4}.---\left(34\right)$

2) work as H _{(r-g) ij}＞V _{(r-g) ij}The time, under the then upper right left side as the edge direction of current point, choice point R _{(i-1) (j+1)}-G _{(i-1) (j+1)}, R _{(i+1) (j-1)}-G _{(i+1) (j-1)}, G _{(i-1) (j+1)}, G _IjAnd G _{(i+1) (j-1)}Calculate current point B _Ij-G _IjThe red green difference R that is lacked _Ij-G _Ij, computational methods are following:

$R_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(R_{(i-1)(j+1)}-G_{(i-1)(j+1)})+(R_{(i+1)(j-1)}-G_{(i+1)(j-1)})}{2}$

$+\frac{(-G_{(i-1)(j+1)}+2G_{\mathrm{ij}}-G_{(i+1)(j-1)})}{4}.---\left(35\right)$

3) work as H _{(r-g) ij}=V _{(r-g) ij}The time, then select R _{(i-1) (j-1)}-G _{(i-1) (j-1)}, R _{(i+1) (j+1)}-G _{(i+1) (j+1)}, R _{(i-1) (j+1)}-G _{(i-1) (j+1)}, R _{(i+1) (j-1)}-G _{(i+1) (j-1)}, G _{(i-1) (j-1)}, G _{(i+1) (j+1)}, G _{(i-1) (j+1)}, G _{(i+1) (j-1)}And G _IjCalculate current point B _Ij-G _IjThe red green difference R that is lacked _Ij-G _Ij, computational methods are following:

$R_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(R_{(i-1)(j-1)}-G_{(i-1)(j-1)})+(R_{(i+1)(j+1)}-G_{(i+1)(j+1)})}{4}$

$+\frac{(R_{(i-1)(j+1)}-G_{(i-1)(j+1)})+(R_{(i+1)(j-1)}-G_{(i+1)(j-1)})}{4}---\left(36\right)$

$+\frac{(-G_{(i-1)(j-1)}-G_{(i-1)(j+1)}+4G_{\mathrm{ij}}-G_{(i+1)(j-1)}-G_{(i+1)(j+1)})}{8}.$

B), only comprising the pixel sampling point of green component G, adjacent on every side has only two red greens on horizontal direction or the vertical direction not good enough, current point is G _Ij

1) when two adjacent red greens almost at current point G _IjHorizontal direction on, promptly put R _{I (j-1)}-G _{I (j-1)}And R _{I (j+1)}-G _{I (j+1)}, current point G then _IjThe red green difference R that is lacked _Ij-G _IjComputational methods be:

$R_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(R_{i(j-1)}-G_{i(j-1)})+(R_{i(j+1)}-G_{i(j+1)})}{2}$

$+\frac{(-G_{i(j-1)}+2G_{\mathrm{ij}}-G_{i(j+1)})}{4}.---\left(37\right)$

2) when two adjacent red greens almost at current point G _IjVertical direction on, i.e. R _{(i-1) j}-G _{(i-1) j}And R _{(i+1) j}-G _{(i+1) j}, current point G then _IjThe red green difference R that is lacked _Ij-G _IjComputational methods be:

$R_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(R_{(i-1)j}-G_{(i-1)j})+(R_{(i+1)j}-G_{(i+1)j})}{2}$

$+\frac{(-G_{(i-1)j}+2G_{\mathrm{ij}}-G_{(i+1)j})}{4}.---\left(38\right)$

Further the reconstruction of full resolution aberration R-G is described and describes below in conjunction with the concrete arrangement mode of Bayer form chromatism data of Fig. 6 and the green component of rebuilding of Fig. 5, so that understand method of the present invention.

A), comprising aberration B-G but the sample point of disappearance aberration R-G; Adjacent on every side has four red greens not good enough; Lay respectively at upper left, left following, bottom right, upper right; Following two miter angle tilted directions are taken all factors into consideration the green component G and the red green difference R-G of current point neighborhood from a upper left bottom right and a upper right left side, calculate upper left lower right respectively to Grad H _R-gWith upper right lower left to Grad V _R-g

With aberration point B ₃₃-G ₃₃Be example, then the gradient H of the 3rd row and the 3rd row _{(r-g) 33}And V _{(r-g) 33}For:

H _(r-g)33＝|(R ₂₂-G ₂₂)-(G ₄₄-G ₄₄)|+|-G ₂₂+2G ₃₃-G ₄₄|. (39)

V _(r-g)33＝|(R ₂₄-G ₂₄)-(R ₄₂-G ₄₂)|+|-G ₂₄+2G ₃₃-G ₄₂|. (40)

Above two Grad are selected the edge direction of less direction as current point.

1) works as G _{(r-g) 33}＜V _{(r-g) 33}The time, then upper left bottom right is as the edge direction of current point, choice point R ₂₂-G ₂₂, R ₄₄-G ₄₄, G ₂₂, G ₃₃And G ₄₄Calculate current point B ₃₃-G ₃₃The red green difference R that is lacked ₃₃-G ₃₃, computational methods are following:

$R_{33}-G_{33}=\frac{(R_{22}-G_{22})+(R_{44}-G_{44})}{2}+\frac{(-G_{22}+2G_{33}-G_{44})}{4}.---\left(41\right)$

2) work as H _{(r-g) 33}＞V _{(r-g) 33}The time, under the then upper right left side as the edge direction of current point, choice point R ₂₄-G ₂₄, R ₄₂-G ₄₂, G ₂₄, G ₃₃And G ₄₂Calculate current point B ₃₃-G ₃₃The red green difference R that is lacked ₃₃-G ₃₃, computational methods are following:

$R_{33}-G_{33}=\frac{(R_{24}-G_{24})+(R_{42}-G_{42})}{2}+\frac{(-G_{24}+2G_{33}-G_{42})}{4}.---\left(42\right)$

3) work as H _{(r-g) 33}=V _{(r-g) 33}The time, then select R ₂₂-G ₂₂, R ₄₄-G ₄₄, R ₂₄-G ₂₄, R ₄₂-G ₄₂, G ₂₂, G ₄₄, G ₂₄, G ₄₂And G ₃₃Calculate current point B ₃₃-G ₃₃The red green difference R that is lacked ₃₃-G ₃₃, computational methods are following:

$R_{33}-G_{33}=\frac{(R_{22}-G_{22})+(R_{44}-G_{44})+(R_{24}-G_{24})+(R_{42}-G_{42})}{4}$

$+\frac{(-G_{22}-G_{24}+4G_{33}-G_{42}-G_{44})}{8}.---\left(43\right)$

With particular point B ₁₅-G ₁₅Be example, then the gradient H of the 1st row and the 5th row _{(r-g) 15}And V _{(r-g) 15}For:

H _(r-g)15＝|(R ₀₄-G ₀₄)|+|G ₁₅-G ₀₄|. (44)

V _(r-g)15＝|(R ₂₄-G ₂₄)|+|G ₁₅-G ₂₄|. (45)

Above two Grad are selected the edge direction of less direction as current point.

1) works as H _{(r-g) 15}＜V _{(r-g) 15}The time, then upper left bottom right is as the edge direction of current point, choice point R ₀₄-G ₀₄, G ₁₅And G ₀₄Calculate current point B ₁₅-G ₁₅The red green difference R that is lacked ₁₅-G ₁₅, computational methods are following:

$R_{15}-G_{15}=(R_{04}-G_{04})+\frac{(G_{15}-G_{04})}{2}.---\left(46\right)$

2) work as H _{(r-g) 15}＞V _{(r-g) 15}The time, under the then upper right left side as the edge direction of current point, choice point R ₂₄-G ₂₄, G ₁₅And G ₂₄Calculate current point B ₁₅-G ₁₅The red green difference R that is lacked ₁₅-G ₁₅, computational methods are following:

$R_{15}-G_{15}=(R_{24}-G_{24})+\frac{(G_{15}-G_{24})}{2}.---\left(47\right)$

3) work as H _{(r-g) 15}=V _{(r-g) 15}The time, then select R ₀₄-G ₀₄, R ₂₄-G ₂₄, G ₀₄, G ₁₅And G ₂₄Calculate current point B ₁₅-G ₁₅The red green difference R that is lacked ₁₅-G ₁₅, computational methods are:

$R_{15}-G_{15}=\frac{(R_{04}-G_{04})+(R_{24}-G_{24})}{2}+\frac{(-G_{04}+2G_{15}-G_{24})}{4}.---\left(48\right)$

B), only comprising the pixel sampling point of green component G, adjacent on every side has only two red greens on horizontal direction or the vertical direction not good enough.

1) with a G ₂₃Be example, two adjacent red greens almost on the horizontal direction of current point, are promptly put R ₂₂-G ₂₂And R ₂₄-G ₂₄, current point G then ₂₃The red green difference R that is lacked ₂₃-G ₂₃Computational methods be:

$R_{23}-G_{23}=\frac{(R_{22}-G_{22})+(R_{24}-G_{24})}{2}+\frac{(-G_{22}+2G_{23}-G_{24})}{4}.---\left(49\right)$

2) with particular point G ₅₂Be example, an adjacent red green almost on the vertical direction of current point, is promptly put R ₄₂-G ₄₂, current point G then ₅₂The red green difference R that is lacked ₅₂-G ₅₂Computational methods be:

$R_{52}-G_{52}=(R_{42}-G_{42})+\frac{(G_{52}-G_{42})}{2}.---\left(50\right)$

Fig. 7 has showed 6 * 6 red green difference R-G that rebuild that calculate through step 3.

(2), the reconstruction of full resolution color difference image B-G

A), comprising aberration R-G but the sample point of disappearance aberration B-G; Adjacent on every side has four blue-greens not good enough; Lay respectively at upper left, left following, bottom right, upper right; Following two miter angle tilted directions are taken all factors into consideration the green component G and the bluish-green aberration B-G of current point neighborhood from a upper left bottom right and a upper right left side, calculate upper left lower right respectively to Grad H _B-gWith upper right lower left to Grad V _B-gCurrent point is R _Ij-G _Ij, the aberration gradient H that then image i is capable and j is listed as _{(b-g) ij}And V _{(b-g) ij}For:

H _(b-g)ij＝|(B _(i-1)(j-1)-G _(i-1)(j-1))-(B _(i+1)(j+1)-G _(i+1)(j+1))|

(51)

+|-G _(i-1)(j-1)+2G _ij-G _(i+1)(j+1)|.

V _(b-g)ij＝|(B _(i-1)(j+1)-G _(i-1)(j+1))-(B _(i+1)(j-1)-G _(i+1)(j-1))|

(52)

+|-G _(i-1)(j+1)+2G _ij-G _(i+1)(j-1)|.

Above two Grad are selected the edge direction of less direction as current point.

1) works as H _{(b-g) ij}＜V _{(b-g) ij}The time, then upper left bottom right is as the edge direction of current point, choice point B _{(i-1) (j-1)}-G _{(i-1) (j-1)}, B _{(i+1) (j+1)}-G _{(i+1) (j+1)}, G _{(i-1) (j-1)}, G _IjAnd G _{(i+1) (j+1)}Calculate current point R _Ij-G _IjThe bluish-green aberration B that is lacked _Ij-G _Ij, computational methods are following:

$B_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(B_{(i-1)(j-1)}-G_{(i-1)(j-1)})+(B_{(i+1)(j+1)}-G_{(i+1)(j+1)})}{2}$

$+\frac{(-G_{(i-1)(j-1)}+2G_{\mathrm{ij}}-G_{(i+1)(j+1)})}{4}.---\left(53\right)$

2) work as H _{(b-g) ij}＞V _{(b-g) ij}The time, under the then upper right left side as the edge direction of current point, choice point B _{(i-1) (j+1)}-G _{(i-1) (j+1)}, B _{(i+1) (j-1)}-G _{(i+1) (j-1)}, G _{(i-1) (j+1)}, G _IjAnd G _{(i+1) (j-1)}Calculate current point R _Ij-G _IjThe bluish-green aberration B that is lacked _Ij-G _Ij, computational methods are following:

$B_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(B_{(i-1)(j+1)}-G_{(i-1)(j+1)})+(B_{(i+1)(j-1)}-G_{(i+1)(j-1)})}{2}$

$+\frac{(-G_{(i-1)(j+1)}+2G_{\mathrm{ij}}-G_{(i+1)(j-1)})}{4}.---\left(54\right)$

3) work as H _{(b-g) ij}=V _{(b-g) ij}The time, then select B _{(i-1) (j-1)}-G _{(i-1) (j-1)}, B _{(i+1) (j+1)}-G _{(i+1) (j+1)}, B _{(i-1) (j+1)}-G _{(i-1) (j+1)}, B _{(i+1) (j-1)}-G _{(i+1) (j-1)}, G _{(i-1) (j-1)}, G _{(i+1) (j+1)}, G _{(i-1) (j+1)}, G _{(i+1) (j-1)}And G _IjCalculate current point R _Ij-G _IjThe bluish-green aberration B that is lacked _Ij-G _Ij, computational methods are following:

$B_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(B_{(i-1)(j-1)}-G_{(i-1)(j-1)})+(B_{(i+1)(j+1)}-G_{(i+1)(j+1)})}{4}$

$+\frac{(B_{(i-1)(j+1)}-G_{(i-1)(j+1)})+(B_{(i+1)(j-1)}-G_{(i+1)(j-1)})}{4}---\left(55\right)$

$+\frac{(-G_{(i-1)(j-1)}-G_{(i-1)(j+1)}+4G_{\mathrm{ij}}-G_{(i+1)(j-1)}-G_{(i+1)(j+1)})}{8}.$

B), only comprising the pixel sampling point of green component G, adjacent on every side has only two blue-greens on horizontal direction or the vertical direction not good enough, current point is G _Ij

1) when two adjacent blue-greens almost at current point G _IjHorizontal direction on, promptly put B _{I (j-1)}-G _{I (j-1)}And B _{I (j+1)}-G _{I (j+1)}, current point G then _IjThe bluish-green aberration B that is lacked _Ij-G _IjComputational methods be:

$B_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(B_{i(j-1)}-G_{i(j-1)})+(B_{i(j+1)}-G_{i(j+1)})}{2}$

$+\frac{(-G_{i(j-1)}+2G_{\mathrm{ij}}-G_{i(j+1)})}{4}.---\left(56\right)$

2) when two adjacent blue-greens almost at current point G _IjVertical direction on, i.e. B _{(i-1) j}-G _{(i-1) j}And B _{(i+1) j}-G _{(i+1) j}, current point G then _IjThe bluish-green aberration B that is lacked _Ij-G _IjComputational methods be:

$B_{\mathrm{ij}}-G_{\mathrm{ij}}=\frac{(B_{(i-1)j}-G_{(i-1)j})+(B_{(i+1)j}-G_{(i+1)j})}{2}$

$+\frac{(-G_{(i-1)j}+2G_{\mathrm{ij}}-G_{(i+1)j})}{4}.---\left(57\right)$

Further the reconstruction of full resolution aberration B-G is described and describes below in conjunction with the concrete arrangement mode of Bayer form chromatism data of Fig. 6 and the green component of rebuilding of Fig. 5, so that understand method of the present invention.

A), comprising aberration R-G but the sample point of disappearance aberration B-G; Adjacent on every side has four blue-greens not good enough; Lay respectively at upper left, left following, bottom right, upper right; Following two miter angle tilted directions are taken all factors into consideration the green component G and the bluish-green aberration B-G of current point neighborhood from a upper left bottom right and a upper right left side, calculate upper left lower right respectively to Grad H _B-gWith upper right lower left to Grad V _B-g

With aberration point R ₂₂-G ₂₂Be example, then the gradient H of the 2nd row and the 2nd row _{(b-g) 22}And V _{(b-g) 22}For:

H _(b-g)22＝|(B ₁₁-G ₁₁)-(B ₃₃-G ₃₃)|+|-G ₁₁+2G ₂₂-G ₃₃|. (58)

V _(b-g)22＝|(B ₁₃-G ₁₃)-(B ₃₁-G ₃₁)|+|-G ₁₃+2G ₂₂-G ₃₁|. (59)

Above two Grad are selected the edge direction of less direction as current point.

1) works as H _{(b-g) 22}＜V _{(b-g) 22}The time, then upper left bottom right is as the edge direction of current point, choice point B ₁₁-G ₁₁, B ₃₃-G ₃₃, G ₁₁, G ₂₂And G ₃₃Calculate current point R ₂₂-G ₂₂The red green difference B that is lacked ₂₂-G ₂₂, computational methods are following:

$B_{22}-G_{22}=\frac{(B_{11}-G_{11})+(B_{33}-G_{33})}{2}+\frac{(-G_{11}+2G_{22}-G_{33})}{4}.---\left(60\right)$

2) work as H _{(b-g) 22}＞V _{(b-g) 22}The time, under the then upper right left side as the edge direction of current point, choice point B ₁₃-G ₁₃, B ₃₁-G ₃₁, G ₁₃, G ₂₂And G ₃₁Calculate current point R ₂₂-G ₂₂The red green difference B that is lacked ₂₂-G ₂₂, computational methods are following:

$B_{22}-G_{22}=\frac{(B_{13}-G_{13})+(B_{31}-G_{31})}{2}+\frac{(-G_{13}+2G_{22}-G_{31})}{4}.---\left(61\right)$

3) work as H _{(b-g) 22}=V _{(b-g) 22}The time, then select B ₁₁-G ₁₁, B ₁₃-G ₁₃, B ₃₁-G ₃₁, B ₃₃-G ₃₃, G ₁₁, G ₃₃, G ₁₃, G ₂₂And G ₃₁Calculate current point R ₂₂-G ₂₂The red green difference B that is lacked ₂₂-G ₂₂, computational methods are following:

$B_{22}-G_{22}=\frac{(B_{11}-G_{11})6(B_{13}-G_{13})+(B_{31}-G_{31})+(B_{33}-G_{33})}{4}$

$+\frac{(-G_{11}-G_{13}+4G_{22}-G_{31}-G_{33})}{8}.---\left(62\right)$

With particular point R ₀₀-G ₀₀Be example, the bluish-green aberration that this point is adjacent has only B ₁₁-G ₁₁, the bluish-green aberration B that lacked of this point then ₀₀-G ₀₀For

$B_{00}-G_{00}=(B_{11}-G_{11})+\frac{(G_{00}-G_{11})}{2}.---\left(63\right)$

B), only comprising the pixel sampling point of green component G, adjacent on every side has only two blue-greens on horizontal direction or the vertical direction not good enough.

1) with a G ₃₂Be example, two adjacent blue-greens almost on the horizontal direction of current point, are promptly put B ₃₁-G ₃₁And B ₃₃-G ₃₃, current point G then ₃₂The bluish-green aberration B that is lacked ₃₂-G ₃₂Computational methods be:

$B_{32}-G_{32}=\frac{(B_{31}-G_{31})+(B_{33}-G_{33})}{2}+\frac{(-G_{31}+2G_{32}-G_{33})}{4}.---\left(64\right)$

2) with a G ₂₃Be example, two adjacent blue-greens almost on the vertical direction of current point, are promptly put B ₁₃-G ₁₃And B ₃₃-G ₃₃, current point G then ₂₃The bluish-green aberration B that is lacked ₂₃-G ₂₃Computational methods be:

$B_{23}-G_{23}=\frac{(B_{13}-G_{13})+(B_{33}-G_{33})}{2}+\frac{(-G_{13}+2G_{23}-G_{33})}{4}.---\left(65\right)$

Fig. 8 has showed 6 * 6 bluish-green aberration B-G that rebuild that calculate through step 3.

The reconstruction of step 4. full resolution red component R and blue component B

On the basis of step 1, obtained the full resolution green component G that rebuild, promptly each pixel sampling point of image all includes green component G.On the basis of step 3, obtained the full resolution color difference image R-G and the B-G that have rebuild, promptly each pixel sampling point of image all includes aberration R-G and B-G.

(1), the reconstruction of full resolution red component R

The image current pixel point includes the aberration R that has rebuild _Ij-G _IjWith green component G _Ij, the red component R that current point red green difference and green component values addition are promptly obtained current point and lacked _Ij

The R-G of the reconstruction color difference image of rebuilding green component and Fig. 7 in conjunction with Fig. 5 describes and describes.G with Fig. 5 ₂₃R with Fig. 7 ₂₃-G ₂₃Be example, then the red component of image the 2nd row and the 3rd row is (R ₂₃-G ₂₃)+G ₂₃

The span of red component R is 0 to 255, therefore when more than the red component R that calculates need overflow compensation deals when exceeding this span, promptly when the result of calculation of R less than 0 the time, the then current red component R that rebuilds is 0; When the result of calculation of R greater than 255 the time, the then current red component R that rebuilds is 255.Fig. 9 has showed 6 * 6 red component R that rebuild that calculate.

(2), the reconstruction of full resolution blue component B

The image current pixel point includes the aberration B that has rebuild _Ij-G _IjWith green component G _Ij, the blue component B that current point blue-green difference and green component values addition are promptly obtained current point and lacked _Ij

The B-G of the reconstruction color difference image of rebuilding green component and Fig. 8 in conjunction with Fig. 5 describes and describes.G with Fig. 5 ₄₂B with Fig. 8 ₄₂-G ₄₂Be example, then the blue component of image the 4th row and the 2nd row is (B ₄₂-G ₄₂)+G ₄₂

The span of blue component B is 0 to 255, therefore when more than the blue component B ultrasonic that calculates need overflow compensation deals when going out this span, promptly when the result of calculation of B less than 0 the time, the then current blue component B that rebuilds is 0; When the result of calculation of B greater than 255 the time, the then current blue component B that rebuilds is 255.Figure 10 has showed 6 * 6 blue component B that rebuild that calculate.

Claims (1)

1. the Bayer form color interpolation method based on green component and aberration space is characterized in that this method comprises the steps:

Step 1. is rebuild full resolution green component G, and concrete grammar is:

At the pixel sampling point place that comprises red component R, take all factors into consideration the green component G and the red component R of current point neighborhood, calculated level direction gradient H from horizontal direction and vertical direction _GrWith vertical gradient V _Gr

1-1. work as H _Gr＞V _GrThe time, then vertical direction selects two adjacent on current point vertical direction green component G and two red component R and current red component R to calculate the green component G that current point lacks as the edge direction of current point;

1-2. work as H _Gr＜V _GrThe time, then horizontal direction selects two adjacent on current point horizontal direction green component G and two red component R and current red component R to calculate the green component G that current point lacks as the edge direction of current point;

1-3. work as H _Gr=V _GrThe time, then select four adjacent green component G of current point and four red component R and current red component R to calculate the green component G that current point lacks on every side;

At the pixel sampling point place that comprises blue component B, take all factors into consideration the green component G and the blue component B of current point neighborhood, calculated level direction gradient H from horizontal direction and vertical direction _GbWith vertical gradient V _Gb, select the edge direction of the less direction of gradient as current point;

1-4. work as H _Gb＞V _GbThe time, then vertical direction selects two adjacent on current point vertical direction green component G and two blue component B and current blue component B to calculate the green component G that current point lacks as the edge direction of current point;

1-5. work as H _Gb＜V _GbThe time, then horizontal direction selects two adjacent on current point horizontal direction green component G and two blue component B and current blue component B to calculate the green component G that current point lacks as the edge direction of current point;

1-6. work as H _Gb=V _GbThe time, then select four adjacent green component G of current point and four blue component B and current blue component B to calculate the green component G that current point lacks on every side;

Step 2. is calculated the pixel aberration R-G that comprises red component and is calculated the pixel aberration B-G that comprises blue component, and concrete grammar is:

Comprising the pixel sampling point of red component R, on the basis of step 1, the green component G that current point includes original red color component R and rebuild, calculated difference R-G obtains the red green difference image of Bayer form;

Comprising the pixel sampling point of blue component B, on the basis of step 1, the green component G that current point includes original blue component B and rebuild, calculated difference B-G obtains the blue-green difference image of Bayer form;

Step 3. is rebuild full resolution color difference image R-G and color difference image B-G, and concrete grammar is:

Rebuild the concrete grammar of full resolution color difference image R-G:

Comprising aberration B-G but the sample point of disappearance aberration R-G; Adjacent on every side has four red greens not good enough; Lay respectively at upper left, left following, bottom right, upper right; Following two miter angle tilted directions are taken all factors into consideration the green component G and the red green difference R-G of current point neighborhood from a upper left bottom right and a upper right left side, calculate upper left lower right respectively to Grad H _R-gWith upper right lower left to Grad V _R-g, select the edge direction of the less direction of Grad as current point;

3-1. work as H _R-g＞V _R-gThe time; As the edge direction of current point, the green component G that two red green difference R-G that select that the upper right lower left of current point makes progress adjacent and two green component G that rebuild and current point have been rebuild calculates the red green difference R-G that current point lacks under the then upper right left side;

3-2. work as H _R-g＜V _R-gThe time; Then upper left bottom right is as the edge direction of current point, and the green component G that two red green difference R-G that select that the upper left lower right of current point makes progress adjacent and two green component G that rebuild and current point have been rebuild calculates the red green difference R-G that current point lacks;

3-3. work as H _R-g=V _R-gThe time, the green component G that then selects four adjacent red green difference R-G of two miter angle tilted directions of current point and four green component G that rebuild and current point to rebuild calculates the red green difference R-G that current point lacks;

At the pixel sampling point place that only comprises green component G; Adjacent on every side has two red greens not good enough; Employing is got mean deviation to two adjacent red green difference R-G and is revised with not good enough green component and the current point green component G that locates of adjacent red green, obtains the red green difference R-G that current point lacks;

Rebuild the concrete grammar of full resolution color difference image B-G:

Comprising aberration R-G but the sample point of disappearance aberration B-G; Adjacent on every side has four red greens not good enough; Lay respectively at upper left, left following, bottom right, upper right; Following two miter angle tilted directions are taken all factors into consideration the green component G and the bluish-green aberration B-G of current point neighborhood from a upper left bottom right and a upper right left side, calculate upper left lower right respectively to Grad H _B-gWith upper right lower left to Grad V _B-g, select the edge direction of the less direction of Grad as current point;

3-4. work as H _B-g＞V _B-gThe time; As the edge direction of current point, the green component G that two red green difference R-G that select that the upper right lower left of current point makes progress adjacent and two green component G that rebuild and current point have been rebuild calculates the red green difference R-G that current point lacks under the then upper right left side;

3-5. work as H _B-g＜V _B-gThe time; Then upper left bottom right is as the edge direction of current point, and the green component G that two red green difference R-G that select that the upper left lower right of current point makes progress adjacent and two green component G that rebuild and current point have been rebuild calculates the red green difference R-G that current point lacks and works as H _B-g＜V _B-gThe time; Then upper left bottom right is as the edge direction of current point, and the green component G that two red green difference R-G that select that the upper left lower right of current point makes progress adjacent and two green component G that rebuild and current point have been rebuild calculates the red green difference R-G that current point lacks;

3-6. work as H _B-g=V _B-gThe time, the green component G that then selects four adjacent red green difference R-G of two miter angle tilted directions of current point and four green component G that rebuild and current point to rebuild calculates the red green difference R-G that current point lacks;

Only comprising the pixel sampling point of green component G; Adjacent on every side has two blue-greens not good enough; Employing is got mean deviation to two adjacent bluish-green aberration B-G and is revised with not good enough green component and the current point green component G that locates of adjacent blue-green, obtains the bluish-green aberration B-G that current point lacks;

Step 4. is rebuild full resolution red component R and blue component B;

The concrete grammar of rebuilding full resolution red component R is:

With red green difference R-G in each pixel sampling point of image and the green component G addition in the current sampling point, calculate all sampled points of image and promptly obtained full resolution red component R, promptly each pixel sampling point of image all includes red component R;

The concrete grammar of rebuilding full resolution blue component B is:

With blue-green difference B-G in each pixel sampling point of image and the green component G addition in the current sampling point, calculate all sampled points of image and promptly obtained full resolution blue component B, promptly each pixel sampling point of image all includes blue component B.

Images