CN111654705B - Mosaic image compression method based on color space conversion - Google Patents

Abstract

The invention belongs to the technical field of image processing, and particularly provides a mosaic image compression method based on novel color space conversion, which is used for solving the problems that the recovered image quality is low, the algorithm is complex and the method is only suitable for RGGB type CFA images with a certain specific format in the prior art. The invention converts G of CFA into G ₁ RBG ₂ The four-channel model is converted, then the novel color space conversion is used, the correlation among pixels is eliminated, and the digit of the coded code stream is greatly reduced, so that the coded code stream can be directly coded through a traditional coding frame; moreover, the invention is applicable to RGGB type CFA of any format; meanwhile, the color space conversion matrix and the inverse conversion matrix are both true values, but not approximate results, so that rounding errors caused by the conversion matrix are eliminated; in addition, the invention updates the brightness component to be coded in a mathematical optimization mode, thereby bringing better recovery quality.

Description

Mosaic image compression method based on color space conversion

technical field

The invention belongs to the technical field of image processing, in particular to the field of RGGB type CFA compression coding, and specifically provides a mosaic image compression method based on novel color space conversion.

Background technique

To produce color images, most digital cameras use a CMOS board with several different color filters and use interpolation techniques to produce full-color images. RGGB color filter array (Color Filter Array, CFA) refers to the general term of CFA with two green color filters, one red color filter and one blue color filter in a 2×2 area; among them, Bayer-CFA is the most commonly used , in each 2×2 region, the upper left and lower right are green color filters, the upper right is red color filter, and the lower left is color filter. If a full-color image is interpolated, there is only one color component in each pixel of the CFA, so the other two color components of a given pixel must be interpolated using neighboring pixel information; although there are several possible interpolation algorithms, very few Obviously, from the point of view of information theory, they all lead to the increase of redundancy.

At present, most of the CMOS image compression methods on civilian digital cameras are image compression after image interpolation, and the amount of compressed data is three times that of CFA, which is not conducive to the real-time performance of space camera image compression. The existing RGGB-type CFA compression method either converts two green pixels into one, which reduces the quality of the restored image; or designs complex structural transformation or numerical filling for the Y matrix, and the algorithm is complex and only applicable to a specific format RGGB-type CFA.

Based on this, the present invention provides a mosaic image compression method based on novel color space conversion.

Contents of the invention

The purpose of the present invention is to provide a mosaic image compression method based on a novel color space conversion for the problems of low image quality, complex algorithm and only applicable to a certain format of RGGB type CFA image in the prior art, which is simple to implement , which can bring higher recovery quality under the condition of the same number of coded stream bits, and is applicable to RGGB type CFA in any format.

To achieve the above object, the technical solution adopted in the present invention is:

A mosaic image compression method based on novel color space conversion, comprising the following steps:

Step 1. The G ₁ , R, B, and G ₂ components in the RGGB-type CFA image are spliced separately to form a channel to generate a four-channel model;

Step 2. Perform color space conversion on the four-channel model generated in step 1, and convert it to the YDgCoCg model to obtain the luminance component Y, chrominance components Dg, Co, and Cg;

Step 3, carry out DCT, JPEG chromatic quantification to the chromaticity component Dg, Co, Cg in the step 2, generate chromaticity component Dg ', Co ', Cg ';

Step 4. Dequantize the chrominance component Cg′ in step 3, generate the chrominance component Cg″ by IDCT, and perform mathematical optimization to calculate and generate the luminance component

进行DCT、JPEG亮度量化，生成亮度分量

Step 5, for the brightness component in step 4

Perform DCT and JPEG luminance quantization to generate luminance components

进行JPEG亮度编码，并将码流发送；Step 6. Carry out JPEG chroma encoding to the chroma components Dg', Co', Cg' in the step 3, and carry out the chroma encoding in the step 5 to the luminance component

Perform JPEG brightness encoding and send the code stream;

Step 7: Receive the code stream, perform inverse coding, inverse quantization, IDCT, inverse color space conversion, and recover the CFA image.

Further, the color space conversion in the step 2 is:

Further, in the step 4, the mathematical optimization is:

Further, the step 7 specifically includes:

Dg″、Co″、Cg″；S71. Receive data, and sequentially perform inverse encoding, inverse quantization, and IDCT to generate components

Dg", Co", Cg";

Dg″、Co″、Cg″进行逆色彩空间转换，得到分量G₁″、R″、B″′、G″₂：S72, paired components

Dg″, Co″, Cg″ perform inverse color space conversion to obtain components G ₁ ″, R″, B″′, G″ ₂ :

S73. Perform the reverse operation on the components G ₁ ″, R″, B″′, and G″ ₂ as in step 1, and recover an RGGB-type CFA image.

The working principle of the present invention is:

The present invention proposes a four-channel model, through color space conversion, the component size of the luma code is changed to 1/4 of the CFA size, and the number of bits of the code stream generated by the chroma code is very small, so the number of bits of the code stream is effectively reduced ;

The color space conversion method is:

Among them, Y is the luminance component, Dg, Co, and Cg are the chrominance components, and Dg represents the green difference: Dg=G ₂ -G ₁ ;

Assuming that Y, Dg, Co, and Cg components are respectively DCT, quantized, and encoded to obtain Y', Dg', Co', and Cg', then inverse encoding, inverse quantization, and IDCT restore Y", Dg", Co", and Cg ″, after inverse color space conversion:

At this time, relative to the original CFA, the resulting distortion is:

To minimize distortion:

Find a singularity, and according to the practical sense, this point is the minimum point:

(而不是Y)时，能够带来更小的整体失真，即更高的恢复质量；因此，本发明在编码端处理的亮度分量为

分量经过DCT、量化、编码得到

后，再经过反编码、反量化、ID CT得到的

与

仍然具有一定差距，但这个差距远远小于直接编码Y、解码生成的Y″与

的差距。It can be seen that compression coding causes severe distortion of the chroma channel, which leads to the brightness component obtained by inverse quantization at the decoding end closer to

(instead of Y), can bring smaller overall distortion, that is, higher restoration quality; therefore, the luminance component processed at the encoding end of the present invention is

The components are obtained through DCT, quantization, and encoding

After that, it is obtained by inverse encoding, inverse quantization, and ID CT

and

There is still a certain gap, but this gap is far smaller than the direct coding Y, the Y″ generated by decoding and the

difference.

In summary, the beneficial effects of the present invention are:

The present invention provides a mosaic image compression method based on a new color space conversion, which converts G ₁ RBG ₂ of CFA into a four-channel model, and then uses a new color space conversion to eliminate the correlation between pixels and greatly reduce the encoding code stream The number of digits, so that it can be directly encoded by the traditional coding framework; and, the present invention is applicable to RGGB type CFA in any format; meanwhile, the color space transformation matrix and inverse transformation matrix in the present invention are real values, rather than approximate results , which eliminates the rounding error caused by the transformation matrix; in addition, the present invention updates the luminance component to be coded through mathematical optimization, resulting in better restoration quality.

Description of drawings

FIG. 1 is a schematic flow chart of the mosaic image compression method based on the novel color space conversion of the present invention.

FIG. 2 is a schematic diagram of the color of each pixel of the Bayer format CFA in the embodiment of the present invention.

FIG. 3 is a schematic diagram of color space conversion in an embodiment of the present invention.

Detailed ways

The present invention will be described in further detail below in conjunction with the accompanying drawings and embodiments.

This embodiment provides a mosaic image compression method based on novel color space conversion, and its process is shown in Figure 1; this embodiment takes Bayer-CFA as an example, and specifically includes the following steps:

Step S1, the G ₁ RBG ₂ components in the RGGB type CFA image are assembled into a channel separately, and become a four-channel model, so that each component forms a matrix separately, eliminating the complex structure of the CFA;

Specifically, as shown in Figure 2, Bayer-CFA divides Bayer-CFA into 2×2 pixel blocks; as shown in Figure 3, select all elements in the upper left corner of 2×2 pixel blocks to form _G1 components, and select all The elements in the upper right corner of the 2×2 pixel block are combined into the R component, the elements in the lower left corner of all the 2×2 pixel blocks are combined into the B component, and the elements in the lower right corner of all the 2×2 pixel blocks are combined into the G ₂ component;

Step S2, the four-channel G ₁ RBG ₂ model is transferred to the four-channel YDgCoCg model, as shown in Figure 3, the correlation between the four channels of G ₁ RBG ₂ is eliminated, and the number of bits required for encoding is reduced;

Step S3, carry out DCT, JPEG chroma quantification to Dg, Co, Cg component in S2, generate Dg', Co', Cg';

恢复出的亮度分量更接近

将会带来更小的失真；Step S4, IDCT, inverse quantization Cg' to generate Cg", mathematical optimization, calculation generation

The restored luminance component is closer to

will result in less distortion;

Specifically, IDCT, inverse quantization Cg' generates Cg", and calculates a new component

进行DCT、JPEG亮度量化，生成

Step S5, to

Carry out DCT, JPEG brightness quantization, generate

进行JPEG亮度编码，并将码流发送；Step S6, carry out JPEG chromaticity coding to Dg ', Co ', Cg ' that generate in S3, to the Dg ' that produces in S5

Perform JPEG brightness encoding and send the code stream;

Step S7, receiving the code stream, performing inverse encoding, inverse quantization, IDCT, inverse color space conversion, and recovering the CFA image; specifically:

Dg″、Co″、Cg″；S71. Receive data, perform inverse encoding, inverse quantization, and IDCT generation in sequence

Dg", Co", Cg";

Dg″、Co″、Cg″进行逆色彩空间转换：S72. Will

Dg″, Co″, Cg″ perform inverse color space conversion:

S73. Perform an operation opposite to that of S1 to recover Bayer-CFA.

The restoration quality of the compressed image is calculated by restoring the CFA image, and the invention has higher restoration quality. In the present invention, the size of the component that requires luminance coding is 1/4 of the CFA size, so the number of bits in the coded stream is less; and the traditional coding framework can be directly used for direct coding, which has high robustness; finally, it needs to be explained Notably, the Bayer format CFA is taken as an example in this embodiment, but the present invention is also valid for any RGGB type CFA.

The above is only a specific embodiment of the present invention. Any feature disclosed in this specification, unless specifically stated, can be replaced by other equivalent or alternative features with similar purposes; all the disclosed features, or All method or process steps may be combined in any way, except for mutually exclusive features and/or steps.

Claims (1)

1. The mosaic image compression method based on color space conversion comprises the following steps:

step 1, G in RGGB type CFA image ₁ 、R、B、G ₂ The components are respectively and independently spliced to form a channel, and a four-channel model is generated;

step 2, performing color space conversion on the four-channel model generated in the step 1, and converting the four-channel model into an YDgCoCg model to obtain a brightness component Y, chrominance components Dg, co and Cg; the color space conversion is specifically:

step 3, performing DCT and JPEG chroma quantization on the chroma components Dg, co and Cg in the step 2 to generate chroma components Dg ', co ' and Cg ';

step 4, inverse quantization is carried out on the chrominance component Cg 'in the step 3, IDCT is carried out to generate chrominance component Cg', mathematical optimization is carried out, and the luminance component is generated through calculation

The mathematical optimization specifically comprises the following steps:

step 5, comparing the brightness component in step 4

DCT and JPEG brightness quantization are performed to generate brightness components

Step 6, JPEG chroma coding is carried out on the chroma components Dg ', co ' and Cg ' in the step 3, and the brightness component in the step 5 is subjected to

JPEG brightness coding is carried out, and the code stream is sent;

step 7, receiving the code stream, performing inverse coding, inverse quantization, IDCT and inverse color space conversion, and recovering the CFA image; the method comprises the following specific steps:

s71, receiving data, and sequentially performing inverse coding, inverse quantization and IDCT to generate components

Dg″、Co″、Cg″；

S72, component pair

Dg ', co ', cg ' are inverse color space converted to obtain a component G ₁ ″、R″、B″′、G″ ₂ ：

S73, component G ₁ ″、R″、B″′、G″ ₂ And (4) performing the reverse operation of the step 1 to recover the RGGB type CFA image.

Images