KR102025494B1 - Apparatus based on visually lossless compression method for compression of bayer image and method therefor - Google Patents

Abstract

The present invention relates to a method for encoding a Bayer image including an RGB channel, which comprises the following steps: (a-1) estimating a value of a previous pixel using a plurality of pixels before a current pixel; (b-1) estimating variation between the plurality of pixels before the current pixel; (c-1) generating a plurality of lookup tables using a previous image frame; (d-1) selecting one of the plurality of lookup tables generated in the step (c-1) by using the variation estimated in the step (b-1); (e-1) subtracting the previous pixel estimated in the step (a-1) from the current pixel, and outputting the difference thereof; and (f-1) receiving a differential value of the step (e-1) to perform fixed length encoding by using the lookup table selected in the step (d-1).

Description

Apparatus and Method Based on Visual Lossless Compression Technique for Bayer Image Compression {APPARATUS BASED ON VISUALLY LOSSLESS COMPRESSION METHOD FOR COMPRESSION OF BAYER IMAGE AND METHOD THEREFOR}

The present invention relates to an apparatus and method based on visual lossless compression for Bayer image compression, and more particularly, to an apparatus and method based on visual lossless compression using adaptive differential pulse code modulation (DPCM). .

Recently, with the development of the camera sensor's resolution UHD, the data, memory, and bandwidth to be processed by the camera ISP have dramatically increased. Especially for modules that require previous frame information among ISP functions (WDR, 3DNR, etc.), external memory is required. As the resolution increases to UHD, the required memory capacity increases and the required bandwidth increases. For this reason, the image processing is performed by changing the ISP memory to a high specification in order to process UHD data. However, when the image processing is performed by changing to the high specification, the manufacturing cost of the camera ISP (Image Signal Processor) increases due to the change of the high specification memory.

To reduce the amount of data that is processed, there are data compression techniques, which include lossless techniques, visually lossless techniques, and lossy techniques. In the conventional lossless technique, there is no data loss, but the complexity of the operation greatly increases the logic size of the ISP. On the other hand, the visual lossless technique has data loss but the logic size is small compared to the lossless technique due to the low computational complexity. In addition, the conventional lossy technique is a type suitable for processing YC or RGB data, and the efficiency is greatly reduced for Bayer data processing. Therefore, the computational complexity is low and a visual lossless technique that can process Bayer data is required.

Disclosure of Invention The present invention aims to solve the above technical problem, and is a visually lossless compression for Bayer image compression based on adaptive DPCM (Differential Pulse Code Modulation) with low logic complexity. The purpose is to provide a device based on the method and a method thereof.

An encoding method for a Bayer image including an RGB channel of the present invention includes: (a-1) estimating a value of a previous pixel using a plurality of pixels before the current pixel; (b-1) estimating the amount of change between the plurality of pixels before the current pixel; And (c-1) generating a plurality of lookup tables using the previous image frame.

In addition, the encoding method of the present invention, (d-1) using the amount of change estimated in the step (b-1), selecting one of a plurality of lookup table generated in the step (c-1); (e-1) differentially outputting the previous pixel estimated in the step (a-1) from the current pixel; And (f-1) performing a fixed length encoding by receiving the difference value of the step (e-1) using the lookup table selected in the step (d-1). Characterized in that.

Preferably, step (a-1) comprises: at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs, to estimate the value of the previous pixel. Specifically, the (a-1) step, the current pixel (G ₀₎ is G, a channel, G the current pixel of the pixel of the channel (G ₀₎ of G _{1-pixel-pixel} of the G-channel which is located immediately before; A G ₂ pixel, which is a pixel located behind a pixel of the G channel of the previous line of the line where the current pixel G ₀ is located, among the pixels of the G channel adjacent to the current pixel G ₀ ; A G ₃ pixel, which is a pixel located in front of a pixel of the G channel of the previous line of the line where the current pixel G ₀ is located, among the pixels of the G channel adjacent to the current pixel G ₀ ; And the G ₁ pixel is located in one pixel of the G channel of the previous line of the line, the pixel which is located in front of the pixel of the adjacent G-channel and the G ₁ pixel G ₄ pixels; using the to estimate the value of the previous pixel . In addition, in the step (a-1), when the current pixel A ₀ is an R channel or a B channel, A _{1 which} is a pixel of the corresponding channel located immediately before the current pixel A ₀ among the pixels of the corresponding channel. pixel; A G ₃ pixel that is a pixel of a G channel adjacent to the current pixel A ₀ , among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; And a G ₄ pixel which is a pixel of a G channel adjacent to the A ₁ pixel among the pixels of the G channel of the previous line of the line where the A ₁ pixel is located.

In addition, the step (b-1) may include at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs. Specifically, the (b-1) step, the current pixel (G ₀₎ is G, a channel, G the current pixel of the pixel of the channel (G ₀₎ of G _{1-pixel-pixel} of the G-channel which is located immediately before; A G ₂ pixel, which is a pixel located behind a pixel of the G channel of the previous line of the line where the current pixel G ₀ is located, among the pixels of the G channel adjacent to the current pixel G ₀ ; A G ₃ pixel, which is a pixel located in front of a pixel of the G channel of the previous line of the line where the current pixel G ₀ is located, among the pixels of the G channel adjacent to the current pixel G ₀ ; Of the G pixel of the _first pixel of the G channel of the previous line of the line is located, the pixel which is located in front of the pixel of the adjacent channel and the G pixel G ₁ G ₄ pixels; A G ₅ pixel which is a pixel of the G channel located immediately before the G ₁ pixel among the pixels of the G channel; A G ₆ pixel that is a pixel of a G channel located immediately after the G ₂ pixels among the pixels of a G channel; And the G pixels ₇ pixels which is located in front of the pixels of the G pixel is in line ₅ G channel of the pixel of the G channel of the previous line, and the adjacent pixels of the G _5; using, estimates the amount of change. In addition, in the step (b-1), when the current pixel A ₀ is an R channel or a B channel, A _{1 which} is a pixel of the corresponding channel located immediately before the current pixel A ₀ among the pixels of the corresponding channel. pixel; An A ₂ pixel which is a pixel of the channel located immediately before the A ₁ pixel among the pixels of the channel; A G ₃ pixel that is a pixel of a G channel adjacent to the current pixel A ₀ , among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; A G ₄ pixel that is a pixel of a G channel adjacent to the A ₁ pixel among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; A G ₇ pixel that is a pixel of a G channel adjacent to the A ₂ pixel among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; A G ₂ pixel which is a pixel of the G channel located immediately after the G ₃ pixels among the pixels of the G channel; And a G ₆ pixel which is a pixel of the G channel located immediately after the G ₂ pixels among the pixels of the G channel.

Preferably, the step (c-1) is performed for each of the pixels constituting the previous frame, using the density of the distribution of the difference value with the adjacent previous pixel located immediately before the estimated number of lookup tables. It characterized in that to generate. Specifically, in the step (c-1), when codes of each of the plurality of lookup tables are less than or equal to a predetermined threshold, a code value is generated by a linear equation, and codes of each of the plurality of lookup tables are predetermined. If the threshold is exceeded, the code value is generated by the quadratic equation.

In the step (d-1), it is determined whether at least one of the plurality of changes estimated in the step (b-1) is equal to or greater than a value of a predetermined position of one of the plurality of lookup tables. It is done. Specifically, in the step (d-1), if at least one of the plurality of changes estimated in the step (b-1) is equal to or greater than a value of a predetermined position of one of the plurality of lookup tables. When a lookup table is selected and at least one of the plurality of changes estimated in the step (b-1) is less than a value of a predetermined position of one lookup table of the plurality of lookup tables, the corresponding one of the plurality of lookup tables is selected. It is preferable to select a lookup table other than the lookup table.

A decoding method for a Bayer image including an RGB channel of the present invention includes: (a-2) estimating a value of a previous pixel using a plurality of pixels before the current pixel; (b-2) estimating the amount of change between the plurality of pixels before the current pixel; And (c-2) selecting one of a plurality of pre-stored lookup tables using the change amount estimated in step (b-2).

In addition, the decoding method of the present invention, (d-2) performing a fixed length decoding of the input data using the look-up table selected in the step (c-2); And (e-2) adding and outputting the previous pixel estimated in the step (a-2) from the current pixel.

Preferably, step (a-2) comprises: at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs, to estimate the value of the previous pixel. Specifically, the (a-2) step, the current pixel (G ₀₎ is G, a channel, G the current pixel of the pixel of the channel (G ₀₎ of G _{1-pixel-pixel} of the G-channel which is located immediately before; The current pixel (G ₀₎ of a pixel in the G channel of the previous line of the line, the current pixel (G ₀₎ and the two adjacent pixels of the pixel of the G channel from the G ₂ pixel located behind; The current pixel is (G ₀₎ of pixels of the G-channel which is located in front of the pixels of the G channel of the previous line of the line is located, the current pixel and the two pixels adjacent to (G _{0) 3} G pixel; And G ₄ pixels of the pixel of the G-channel which is located in front of the two pixels of the pixel of the G channel of the previous line of the line is the G ₁ pixels located adjacent to the G ₁ pixels; using a, the value of the previous pixel Estimate. In addition, in the step (a-2), when the current pixel A ₀ is an R channel or a B channel, A ₁ , which is a pixel of the corresponding channel located immediately before the current pixel A ₀ , among the pixels of the corresponding channel. pixel; A G ₃ pixel that is a pixel of a G channel adjacent to the current pixel A ₀ , among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; And a G ₄ pixel which is a pixel of a G channel adjacent to the A ₁ pixel among the pixels of the G channel of the previous line of the line where the A ₁ pixel is located.

In addition, the step (b-2) may include at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs. Specifically, the (b-2) stage, the current pixel (G ₀₎ is G, a channel, G the current pixel of the pixel of the channel (G ₀₎ of G _{1-pixel-pixel} of the G-channel which is located immediately before; The current pixel (G ₀₎ of a pixel in the G channel of the previous line of the line, the current pixel (G ₀₎ and the two adjacent pixels of the pixel of the G channel from the G ₂ pixel located behind; The current pixel is (G ₀₎ of pixels of the G-channel which is located in front of the pixels of the G channel of the previous line of the line is located, the current pixel and the two pixels adjacent to (G _{0) 3} G pixel; Wherein the G ₁ pixels among the pixels of the G channel of the previous line of the line is located, the pixels of the G-channel which is located in front of the two pixels adjacent to the pixel G ₁ G ₄ pixels; A G ₅ pixel which is a pixel of the G channel located immediately before the G ₁ pixel among the pixels of the G channel; A G ₆ pixel that is a pixel of a G channel located immediately after the G ₂ pixels among the pixels of a G channel; ₇ and G pixel is the pixel of the G-channel which is located in front of the two pixels of the pixel of the G channel, and the adjacent G ₅ pixels in the previous line of the line is the G pixels located _5; using, estimates the amount of change. In addition, in the step (b-2), when the current pixel A ₀ is an R channel or a B channel, A ₁ , which is a pixel of the corresponding channel located immediately before the current pixel A ₀ , among the pixels of the corresponding channel. pixel; An A ₂ pixel which is a pixel of the channel located immediately before the A ₁ pixel among the pixels of the channel; A G ₃ pixel that is a pixel of a G channel adjacent to the current pixel A ₀ , among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; A G ₄ pixel that is a pixel of a G channel adjacent to the A ₁ pixel among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; A G ₇ pixel that is a pixel of a G channel adjacent to the A ₂ pixel among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; A G ₂ pixel which is a pixel of the G channel located immediately after the G ₃ pixels among the pixels of the G channel; And a G ₆ pixel which is a pixel of the G channel located immediately after the G ₂ pixels among the pixels of the G channel.

Preferably, the step (c-2) may include determining whether at least one of the plurality of changes estimated in the step (b-2) is equal to or greater than a value of a predetermined position of one of the plurality of lookup tables. It is characterized by. Specifically, in the step (c-2), if at least one of the plurality of changes estimated in the step (b-2) is equal to or greater than a value of a predetermined position of one of the plurality of lookup tables. Selecting a lookup table, and when at least one of the plurality of changes estimated in step (b-2) is less than a value of a predetermined position of one of the plurality of lookup tables, the corresponding one of the plurality of lookup tables; Select a lookup table other than the lookup table.

According to the present invention, an apparatus and a method based on a visually lossless compression technique for Bayer image compression based on adaptive DPCM (Differential Pulse Code Modulation) having low logic complexity can be provided.

1 is a block diagram of an ISP system to which the visual lossless compression technique of the present invention is applied.
2 is a block diagram of an encoding apparatus for a visual lossless compression method according to an embodiment of the present invention.
3 is a block diagram of a decoding apparatus for a visual lossless compression method according to an embodiment of the present invention.
4 is an explanatory diagram for a method of estimating a previous pixel of a G channel.
5 is an explanatory diagram for a method of estimating a previous pixel of an R channel or a B channel;
6 is an explanatory diagram for explaining a method for estimating variation between previous pixels of a G channel.
7 is an explanatory diagram for explaining a method for estimating variation between previous pixels of an R channel or a G channel;
8 is a graph of changes in the lookup table according to the distribution of differential histogram ham.
9 and 10 are exemplary diagrams when the lookup table uses Table 2 (Tb2) and Table 1 (Tb1), respectively.

Hereinafter, an apparatus and a method based on a visual lossless compression technique for Bayer image compression according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

The following examples of the present invention are intended to embody the present invention, but not to limit or limit the scope of the present invention. From the detailed description and examples of the present invention, those skilled in the art to which the present invention pertains can easily be interpreted as belonging to the scope of the present invention.

The present invention proposes an adaptive DPCM (Visually Lossless Compression) technique for efficiently compressing Bayer images. As the camera sensor develops into UHD, the external memory capacity and bandwidth consumed by the camera image signal processor (ISP) are greatly increased. Therefore, the above problem is solved by designing a visual lossless compression encoding apparatus 210 and a decoding apparatus 220 based on input / output of an external memory controller. Conventional-based visual lossless compression technique is performed in YC or RGB environment, but in case of camera ISP, VLC should also be performed in Bayer environment because there is much image processing in Bayer environment. In addition, since the logic size of the conventional visual lossless compression module may be a factor to increase the production cost, it is designed to minimize the internal SRAM usage in consideration of this. In view of this point, the present invention proposes a low complexity adaptive DPCM-based visual lossless compression scheme for Bayer compression.

That is, the main features of the present invention are as follows. In the present invention, in order to solve the problem of increasing the memory capacity and bandwidth of the camera ISP according to the increase of the camera resolution, an adaptive DPCM-based visual lossless compression technique with low logic complexity is developed. The camera ISP is based on Bayer compression because it has a lot of Bayer data processing. In consideration of Bayer compression, the existing Low Complexity Lossless Compression for Images (LOCO-I) scheme is applied to the first predictor 211 or the second predictor 221, and the DPCM is performed after estimating the change amount between the previous G channels. When quantizing the difference data obtained, the table is adaptively selected. In addition, in the case of variable length coding (Variable Length Coding-1), since the complexity of the memory controller is greatly increased, coding is performed using fixed length coding. The present invention can minimize the increase in manufacturing cost by adding a simple based visual lossless compression module when developing a high resolution camera ISP.

In the present invention, in consideration of the complexity of the external memory controller, the fixed length method is applied instead of the variable length method.

The visual lossless technique of the present invention uses an adaptive DPCM scheme considering that the association between adjacent pixels is poor because it is the compression of the Bayer image. First, the first predictor 211 or the second predictor 221 estimates previous data by applying a modified LOCO-I method in consideration of the fact that the input data is a Bayer image. Then, the difference between the estimated data and the data of the current pixel is estimated by the amount of change between the previous data and fixed-length encoded to the code value of the selected lookup table.

The present invention improves memory capacity and bandwidth problems according to the resolution of the camera sensor through a visual lossless compression technique of Bayer image. Finally, it is a compression technique that can not feel the deterioration of quality with the human eye, and is similar in terms of the uncompressed result and subjective image quality, but does not require the use of high-definition external memory, so that the camera ISP manufacturing cost does not increase much. Can be. Even if a high specification external memory is used, it is expected that a high specification UHD ISP can be produced using the memory obtained through compression.

1 is a block diagram of an ISP system 100 to which the visual lossless compression technique of the present invention is applied. Among the functions of the ISP, a memory 170 such as DDR is required in the case of a wide dynamic range (WDR) or a digital noise reduction (DNR) that requires a frame memory. At this time, if the resolution of the sensor increases, the capacity and bandwidth of the memory 170 also increase. Recently, as the camera sensor is developed into UHD, the capacity and bandwidth of the required memory 170 are increased, thereby increasing the production cost of the ISP. Accordingly, in the present invention, as shown in FIG. 1, the encoding device 210 and the decoding device 220 by the visual lossless compression technique are added to the input and output of the module requiring the memory 170, that is, the front of the memory controller 180. Capacity and bandwidth of 170 can be secured.

For reference, as can be seen from FIG. 1, the general ISP system 100 includes a lens shading unit 110, a defect correction unit 120, and a WDR unit. 130, the DNR unit 140, the demosaicing unit 150, and the color toning unit 160 outputs the HDMI.

Examples of the apparatus based on the visual lossless compression technique for compressing the Bayer image may include the encoding apparatus 210 and the decoding apparatus 220.

Hereinafter, the encoding device 210 and the decoding device 220 of the present invention will be described in detail.

The encoding device 210 and the decoding device 220 of the present invention may be implemented by at least a part of a processor such as MCU, DSP, CPU, etc. in the form of a computer program.

2 is a block diagram of an encoding apparatus 210 for a visual lossless compression method according to an exemplary embodiment of the present invention. The encoding apparatus 210 of the present invention has been modified to perform Bayer compression in the existing DPCM scheme.

As can be seen from FIG. 2, the encoding device 210 of the present invention includes a first predictor 211, a first table generator 212, a first table selector 213, and a first difference value output device 214. And a first adder 215 and a first encoder 216.

The first predictor 211 predicts a previous pixel by using correlation between the same channels in consideration of the fact that the input image has a Bayer structure. In addition, the first predictor 211 estimates an edge strength of the current position by measuring a change amount of previous data. That is, the first predictor 211 estimates the value of a previous pixel similar to the current pixel by using the plurality of pixels before the current pixel, and estimates the amount of change between the plurality of pixels before the current pixel. .

The first table generator 212 analyzes the data of the previous image to generate a lookup table suitable for the current frame scene. That is, the first table generator 212 generates a plurality of lookup tables for the current image frame by using the previous image frame.

The first table selector 213 determines the lookup table to be used by the first encoder 216 using the edge strength predicted from the first predictor 211. That is, the first table selector 213 selects one of a plurality of lookup tables generated by the first table generator 212 using the estimated change amount.

The first difference value output unit 214 serves to differentially output the previous pixel estimated by the first predictor 211 from the current pixel.

The first adder 215 adds the output of the first predictor 211 and the output of the first difference value output unit 214 to input the first predictor 211.

The first encoder 216 codes the absolute value of the given difference data and the determined lookup table value into a lookup table value having the least error. In this way, a fixed length of encoded data can be obtained. The encoded data has a sign bit and a table code value. That is, the first encoder 216 receives the difference value of the first difference value output unit 214 using the lookup table selected by the first table selector 213 to perform fixed length encoding. Play a role.

3 is a block diagram of a decoding apparatus 220 for a visual lossless compression method according to an exemplary embodiment of the present invention. The decoding apparatus 220 of the present invention has been modified to restore Bayer compressed data in the existing DPCM scheme.

As can be seen from FIG. 3, the decoding device 220 of the present invention includes a second predictor 221, a second table selector 222, a second decoder 223, and a second adder 224. .

In the decoding device 220 according to the present invention, the second table selector 222 has the same look-up table information as the encoding device 210 of the present invention and the amount of change in the peripheral information obtained by the second predictor 221. Get the code value of the lookup table and determine the sign bit. Then, the previous data determined from the second predictor 221 and the second adder 224 are added. Through this process, a restored Bayer image can be obtained.

In summary, the second predictor 221 estimates a value of a previous pixel that is similar to the current pixel by using a plurality of pixels before the current pixel, and estimates a change amount between the plurality of pixels before the current pixel. do. In addition, in the second table selector 222, a plurality of lookups generated by the first table generator 212 of the encoding apparatus 210 and stored in advance for sharing using the amount of change estimated by the second predictor 221. It selects one of the tables.

In addition, the second decoder 223 performs fixed length decoding of the input data by using the lookup table selected by the second table selector 222. In addition, the second adder 224 adds and outputs the previous pixel estimated by the second predictor 221 from the current pixel.

Hereinafter, the main components of the encoding device 210 and the decoding device 220 of the present invention will be described in more detail.

First  Predictor 211, 2nd  Predictor (221)

The first predictor 211 or the second predictor 221 is a low complexity lossless image compression algorithm (LOCO-I) for low computational complexity (Reference: The LOCO-I lossless image compression algorithm: principles and standardization into JPEG-LS, TIP 2000) is adapted to the Bayer environment. The first predictor 211 or the second predictor 221 performs two functions. First, the first predictor 211 or the second predictor 221 includes a function of estimating a previous value for obtaining a difference from a current value and a transfer used by the first table selector 213 or the second table selector 222. There is a function to estimate the amount of change in the value.

The method of estimating the previous pixel is as follows. In addition, since the Bayer environment, the previous pixel is estimated by different methods for each RGB channel.

4 is an explanatory diagram for a method of estimating a previous pixel of a G channel.

이고

이다.In the case of the G channel, Equation 1 and Equation 2 are used to estimate. here

ego

to be.

,

다. α는 현재 픽셀과 변화량 사이의 거리를 보상하기 위한 가중치이다. D ₁ and D ₂ are changes in the previous data as follows.

,

All. α is a weight for compensating the distance between the current pixel and the amount of change.

As can be seen from [Equation 1] and [Equation 2], in the case of the G channel, the pixel of the same G channel before the current pixel G ₀ in the current line in consideration of the directionality of the horizontal and vertical edges. which estimates the transfer of pixel ₁ is similar to the G pixels and the current pixel in the previous line (G ₀₎ of the current pixel G ₂ 'pixel of pixels adjacent to the other channel (G _0). If the estimated value considering the directionality of the horizontal and vertical edges does not satisfy a certain condition, the G ₁ pixel, G ₂ 'pixel before the current pixel G ₀ in the current line By using the G ₃ 'pixel, which is a pixel of the same channel as the G ₂ ' pixel, the previous pixel similar to the current pixel G ₀ is estimated.

5 is an explanatory diagram for a method of estimating a previous pixel of an R channel or a B channel.

In the present invention, even when the current pixel A ₀ is an R channel or a B channel, only the G channel of the previous line is considered by the first predictor 211 or the second predictor 221 in order to minimize memory usage. Therefore, in case of the R channel, Equation 3 is estimated, and in case of the B channel, Equation 4 is used. Β and γ are weights for compensating for the amount of change in the G channel, which is a different channel from the current pixel.

As can be seen from [Equation 3] and [Equation 4], in the case of the R channel or the B channel, the same channel before the current pixel A ₀ in the current line in consideration of the directionality of the horizontal and vertical edges. of the pixel a ₁ of the pixel in the previous line of pixels in the other channel adjacent to the current pixel (a ₀₎ G ₃ pixels to estimate the previous pixel is similar to the current pixel (a _0). The estimated value at the time of considering the direction of the horizontal or vertical edge does not satisfy the predetermined condition, the previous same pixel in the channel A _1-pixel, G ₃ pixels prior to the current pixel (A ₀₎ in the current line G ₃ Using the G ₄ pixel, which is the pixel of the same channel as the pixel, the previous pixel similar to the current pixel A ₀ is estimated.

That is, the first predictor 211 or the second predictor 221 may include at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs, to estimate the value of the previous pixel.

In detail, when the current pixel G ₀ is a G channel, the first predictor 211 or the second predictor 221 is a pixel of the G channel that is located immediately before the current pixel G ₀ of the pixels of the G channel. G ₁ pixel; Of the current pixel (G ₀₎ of the pixel which is located behind the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of the line in G ₂ pixels; Of the current pixel (G ₀₎ of pixels which is located in front of the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of pixels in the line G _3; G ₁ and G pixels are pixels of ₄ pixels which is located in front of the pixels of the G channel of the previous line adjacent to the pixel of G ₁ in the line; using, characterized in that for estimating the value of the previous pixel. In addition, when the current pixel A ₀ is an R channel or a B channel, an A ₁ pixel which is a pixel of the corresponding channel located immediately before the current pixel A ₀ among the pixels of the corresponding channel; A G ₃ pixel that is a pixel of the G channel adjacent to the current pixel A ₀ of the previous line of the line where the current pixel A ₀ is located; And a G ₄ pixel which is a pixel of a G channel adjacent to the A ₁ pixel of the previous line of the line where the A ₁ pixel is located, it is preferable to estimate the value of the previous pixel.

The estimation of the change amount of the previous value in the first predictor 211 or the second predictor 221 is also applied differently for each channel.

6 is an explanatory diagram for explaining a method for estimating variation between previous pixels of a G channel.

은 이전 픽셀 간의 변화량을 나타낸다.The estimation of the change amount between the previous pixels of the G channel may be expressed as shown in [Equation 5]. In [Equation 5]

Represents the amount of change between the previous pixels.

7 is an explanatory diagram illustrating a method of estimating the amount of change between previous pixels of an R channel or a G channel.

,

는 이전 픽셀 간의 변화량을 나타낸다. As shown in [Equation 6], the R channel estimates the change amount of the same channel among the previous pixels and the change amount of the G channel as shown in [Equation 7].

,

Represents the amount of change between the previous pixels.

That is, the first predictor 211 or the second predictor 221 may include at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs.

In detail, when the current pixel G ₀ is the G channel, the first predictor 211 or the second predictor 221 may include a pixel of the G channel that is located immediately before the current pixel G ₀ of the pixels of the G channel. G ₁ pixel; Of the current pixel (G ₀₎ of the pixel which is located behind the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of the line in G ₂ pixels; Of the current pixel (G ₀₎ of pixels which is located in front of the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of pixels in the line G _3; The G ₁ pixel in the pixel which is located in front of the pixels of the G channel adjacent to the pixel of the previous line of the line G ₁ G ₄ pixels; A G ₅ pixel that is a pixel of the G channel located immediately before the G ₁ pixel among the pixels of the G channel; A G ₆ pixel which is a pixel of the G channel located immediately after the G ₂ pixels among the pixels of the G channel; G ₅ and G pixel is a pixel in the pixel ₇ which is located in front of the pixels of the G channel ₅ adjacent to the G pixel in the previous line of the line located; using, characterized in that for estimating the amount of change.

In addition, when the current pixel A ₀ is an R channel or a B channel, the first predictor 211 or the second predictor 221 includes a corresponding channel located immediately before the current pixel A ₀ among pixels of the corresponding channel. A ₁ pixel which is a pixel of; An A ₂ pixel which is a pixel of the channel located immediately before A ₁ pixel among the pixels of the channel; A G ₃ pixel that is a pixel of a G channel adjacent to the current pixel of the previous line of the line where the current pixel A ₀ is located; A G ₄ pixel that is a pixel of a G channel adjacent to the A ₁ pixel of the previous line of the line where the current pixel A ₀ is located; A G ₇ pixel that is a pixel of a G channel adjacent to the A ₂ pixel of the previous line of the line where the current pixel A ₀ is located; A G ₂ pixel that is a pixel of the G channel located immediately after G ₃ pixels among the pixels of the G channel; And a G ₆ pixel which is a pixel of the G channel located immediately after the G ₂ pixels among the pixels of the G channel.

First  Table generator (212)

The first table generator 212 analyzes a previous scene to generate a lookup table suitable for the current scene. The way of analyzing the previous scene is determined by looking at the histogram of the difference value.

8 shows a graph of a change in the lookup table according to the distribution of differential histogram ham.

As shown in FIG. 8, the code value is taken densely, and the code value is wider when the code value is wider. With this concept, we create a table suitable for the current scene. Lookup tables are generated based on linear and quadratic equations. The generation method of Table 2 (Tb2) is shown in [Equation 8], and the generation method of Table 1 (Tb1) is shown in [Equation 9]. Here, x is a code and y is a code value. th1, th2 are threshold values for determining the dense area, max_dif means the maximum difference value of the previous image, and mid_dif is a predetermined value of the cumulative distribution of the difference histogram of the previous image, for example, the cumulative distribution is 85% It means the difference value of the point.

임.)(here,

being.)

임.)(here,

being.)

That is, the first table generator 212 generates a plurality of lookup tables for each pixel constituting the previous frame by using the density of the distribution of the difference value with the adjacent previous pixel positioned immediately before the estimated one. It is characterized by. In detail, the first table generator 212 generates a code value by a linear equation when a code of each of the plurality of lookup tables is less than or equal to a predetermined threshold, and codes of each of the plurality of lookup tables exceed a predetermined threshold. The encoding method, characterized in that for generating a code value by a quadratic equation.

First  Table selector (213) or 2nd  Table Selector (222)

,

을 이용하여 최적의 룩업 테이블을 결정한다. The first table selector 213 or the second table selector 222 is a change amount between previous data obtained from the first predictor 211 or the second predictor 221.

,

Determine the optimal lookup table using.

The selection of the lookup table by the first table selector 213 or the second table selector 222 may be expressed as shown in [Equation 10].

9 and 10 are exemplary diagrams in the case where Table 2 (Tb2) and Table 1 (Tb1) are used as the lookup table, respectively.

의 변화량이 Tb1의 15번 값보다 크면 현재 위치에서 인코딩 할 때 Tb2를 사용하여 코딩한다. 이를 통해 에지 영역이나 평탄 영역에서 효과적으로 압축이 가능하다.In other words,

If the amount of change is greater than the value of Tb1, it is coded using Tb2 when encoding at the current position. This enables effective compression in the edge or flat areas.

That is, in the first table selector 213 or the second table selector 222, at least one of the plurality of changes estimated by the first predictor 211 or the second predictor 221 may be one of the plurality of lookup tables. It is characterized by determining whether or not a value of a predetermined position of the lookup table.

In more detail, in the first table selector 213 or the second table selector 222, at least one of the plurality of changes estimated by the first predictor 211 or the second predictor 221 may be one of the plurality of lookup tables. If the lookup table is equal to or larger than a predetermined value, one corresponding lookup table is selected, and at least one of the plurality of changes estimated by the first predictor 211 or the second predictor 221 is one of the lookup tables. When the value is less than a predetermined position of the table, a lookup table other than the corresponding one lookup table is selected from among the plurality of lookup tables.

First  Encoder 216 or 2nd  Decoder (223)

The first encoder 216 encodes the lookup table information obtained by the first table selector 213 into a predetermined length. In addition, the second decoder 223 decodes the lookup table information obtained by the second table selector 222 to a predetermined length. Fixed-length coding has the advantage of low complexity and the same compression performance in any scene.

Hereinafter, an encoding method for a Bayer image including an RGB channel according to an exemplary embodiment of the present invention will be described.

Since the encoding method for the Bayer image including the RGB channel according to the preferred embodiment of the present invention uses the above-described encoding apparatus 210, the encoding apparatus 210 includes all the features of the encoding apparatus 210 without further description. Of course. The encoding method of the present invention may be implemented by at least part of a processor such as MCU, DSP, CPU, etc. in the form of a computer program.

The encoding method of the present invention includes estimating a value of a previous pixel similar to the current pixel by using a plurality of pixels before the current pixel (S110); Estimating an amount of change between the plurality of pixels before the current pixel (S120); Generating a plurality of lookup tables for the current image frame by using the previous image frame (S130); Selecting one of a plurality of lookup tables generated in step S130 by using the change amount estimated in step S120 (S140); Differentially outputting the previous pixel estimated in step S110 from the current pixel (S150); And performing a fixed length encoding by receiving the difference value of step S150 using the lookup table selected in step S140 (S160).

In more detail, operation S110 may include at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs, to estimate the value of the previous pixel.

In other words, step S110, the current pixel (G ₀₎ is G, a channel, G-channel of the current pixel of the pixel (G ₀₎ of the pixel G ₁ pixels of the G-channel which is located immediately before; Of the current pixel (G ₀₎ of the pixel which is located behind the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of the line in G ₂ pixels; Of the current pixel (G ₀₎ of pixels which is located in front of the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of pixels in the line G _3; G ₁ and G ₄ pixels of the pixel located in front of the pixels of the G channel adjacent to the pixel G ₁ of the previous line of the line of pixels is located; using the to estimate the value of the previous pixel.

In addition, in operation S110, when the current pixel A ₀ is an R channel or a B channel, an A ₁ pixel which is a pixel of the corresponding channel located immediately before the current pixel A ₀ among the pixels of the corresponding channel; A G ₃ pixel that is a pixel of the G channel adjacent to the current pixel A ₀ of the previous line of the line where the current pixel A ₀ is located; And a G ₄ pixel, which is a pixel of a G channel adjacent to the A ₁ pixel of the previous line of the line where the A ₁ pixel is located, to estimate the value of the previous pixel.

Preferably, step S120 includes: at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs.

In other words, step S120, the current pixel (G ₀₎ is G, a channel, G-channel of the current pixel of the pixel (G ₀₎ of the pixel G ₁ pixels of the G-channel which is located immediately before; Of the current pixel (G ₀₎ of the pixel which is located behind the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of the line in G ₂ pixels; Of the current pixel (G ₀₎ of pixels which is located in front of the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of pixels in the line G _3; The G ₁ pixel in the pixel which is located in front of the pixels of the G channel adjacent to the pixel of the previous line of the line G ₁ G ₄ pixels; A G ₅ pixel that is a pixel of the G channel located immediately before the G ₁ pixel among the pixels of the G channel; A G ₆ pixel which is a pixel of the G channel located immediately after the G ₂ pixels among the pixels of the G channel; G ₅ and G pixel is a pixel in the pixel ₇ which is located in front of the pixels of the G channel ₅ adjacent to the G pixel in the previous line in the line; using, it estimates the amount of change.

In addition, in operation S120, when the current pixel A ₀ is an R channel or a B channel, an A ₁ pixel that is a pixel of a corresponding channel positioned immediately before the current pixel A ₀ among pixels of the corresponding channel; An A ₂ pixel which is a pixel of the channel located immediately before A ₁ pixel among the pixels of the channel; A G ₃ pixel that is a pixel of the G channel adjacent to the current pixel A ₀ of the previous line of the line where the current pixel A ₀ is located; A G ₄ pixel that is a pixel of a G channel adjacent to the A ₁ pixel of the previous line of the line where the current pixel A ₀ is located; A G ₇ pixel that is a pixel of a G channel adjacent to the A ₂ pixel of the previous line of the line where the current pixel A ₀ is located; A G ₂ pixel that is a pixel of the G channel located immediately after G ₃ pixels among the pixels of the G channel; And a G ₆ pixel which is a pixel of the G channel located immediately after the G ₂ pixels among the pixels of the G channel.

In operation S130, a plurality of lookup tables may be generated for each pixel constituting the previous frame by using the density of the distribution of the difference value with the adjacent previous pixel positioned immediately before the estimated value. That is, in step S130, when the codes of the plurality of lookup tables are less than or equal to a predetermined threshold, the code values are generated by the first equation, and when the codes of each of the plurality of lookup tables exceed the predetermined threshold, the second order is generated. It is desirable to generate code values by equations.

In addition, the step S140, characterized in that it is determined whether at least one of the plurality of changes estimated in the step S120 is equal to or greater than a value of a predetermined position of one lookup table of the plurality of lookup tables. In detail, in step S140, when at least one of the plurality of changes estimated in step S120 is equal to or greater than a value of a predetermined position of one lookup table among the plurality of lookup tables, the corresponding lookup table is selected, and in step S120, When at least one of the amounts of changes is less than a value of a predetermined position of one lookup table of the plurality of lookup tables, it is preferable to select a lookup table other than the corresponding one lookup table among the plurality of lookup tables.

Hereinafter, a decoding method for a Bayer image including an RGB channel according to an exemplary embodiment of the present invention will be described.

The decoding method for the Bayer image including the RGB channel according to the preferred embodiment of the present invention uses the above-described encoding apparatus 210, and thus includes all the features of the decoding apparatus 220 without further explanation. Of course. The decoding method of the present invention may be implemented by at least a part of a processor such as MCU, DSP, CPU, etc. in the form of a computer program.

The decoding method of the present invention includes estimating a value of a previous pixel similar to the current pixel by using a plurality of pixels before the current pixel (S210); Estimating an amount of change between the plurality of pixels before the current pixel (S220); Selecting one of a plurality of previously stored lookup tables by using the change amount estimated in step S220 (S230); Performing fixed length decoding of the input data using the lookup table selected in step S230 (S240); And adding and outputting the previous pixel estimated in step S210 from the current pixel (S250).

Step S210 may include at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs, to estimate the value of the previous pixel.

In other words, step S210, the current pixel (G ₀₎ is G, a channel, G-channel of the current pixel of the pixel (G ₀₎ of the pixel G ₁ pixels of the G-channel which is located immediately before; Of the current pixel (G ₀₎ of the pixel which is located behind the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of the line in G ₂ pixels; Of the current pixel (G ₀₎ of pixels which is located in front of the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of pixels in the line G _3; G ₁ and G pixels are pixels of ₄ pixels which is located in front of the pixels of the G channel of the previous line adjacent to the pixel of G ₁ in the line; using, it is preferable to estimate the value of the previous pixel.

In addition, in step S210, when the current pixel A ₀ is an R channel or a B channel, an A ₁ pixel which is a pixel of the corresponding channel located immediately before the current pixel A ₀ among the pixels of the corresponding channel; A G ₃ pixel that is a pixel of the G channel adjacent to the current pixel A ₀ of the previous line of the line where the current pixel A ₀ is located; And a G ₄ pixel, which is a pixel of a G channel adjacent to the A ₁ pixel of the previous line of the line where the A ₁ pixel is located, to estimate the value of the previous pixel.

Step S220 may include at least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And at least one pixel belonging to the G channel of the previous line of the line to which the current pixel belongs.

In other words, step S220, the current pixel (G ₀₎ is G, a channel, G-channel of the current pixel of the pixel (G ₀₎ of the pixel G ₁ pixels of the G-channel which is located immediately before; Of the current pixel (G ₀₎ of the pixel which is located behind the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of the line in G ₂ pixels; Of the current pixel (G ₀₎ of pixels which is located in front of the pixel of the current pixel (G ₀₎ and the adjacent G channel of the previous line of pixels in the line G _3; The G ₁ pixel in the pixel which is located in front of the pixels of the G channel adjacent to the pixel of the previous line of the line G ₁ G ₄ pixels; A G ₅ pixel that is a pixel of the G channel located immediately before the G ₁ pixel among the pixels of the G channel; A G ₆ pixel which is a pixel of the G channel located immediately after the G ₂ pixels among the pixels of the G channel; G ₅ and G pixel is a pixel in the pixel ₇ which is located in front of the pixels of the G channel ₅ adjacent to the G pixel in the previous line in the line; using, it estimates the amount of change.

In addition, in operation S220, when the current pixel A ₀ is an R channel or a B channel, an A ₁ pixel which is a pixel of a corresponding channel located immediately before the current pixel A ₀ among pixels of the corresponding channel; An A ₂ pixel which is a pixel of the channel located immediately before A ₁ pixel among the pixels of the channel; A G ₃ pixel that is a pixel of the G channel adjacent to the current pixel A ₀ of the previous line of the line where the current pixel A ₀ is located; A G ₄ pixel that is a pixel of a G channel adjacent to the A ₁ pixel of the previous line of the line where the current pixel A ₀ is located; A G ₇ pixel that is a pixel of a G channel adjacent to the A ₂ pixel of the previous line of the line where the current pixel A ₀ is located; A G ₂ pixel that is a pixel of the G channel located immediately after G ₃ pixels among the pixels of the G channel; And a G ₆ pixel which is a pixel of the G channel located immediately after the G ₂ pixels among the pixels of the G channel, it is preferable to estimate the amount of change.

In operation S230, it is determined whether at least one of the plurality of changes estimated in operation S220 is equal to or greater than a value of a predetermined position of one of the plurality of lookup tables.

In detail, in step S230, when at least one of the plurality of changes estimated in step S220 is equal to or greater than a value of a predetermined position of one lookup table among the plurality of lookup tables, the corresponding one lookup table is selected, and the number estimated in step S220 is selected. When at least one of the amount of change is less than a value of a predetermined position of one of the lookup tables of the plurality of lookup tables, a lookup table other than the corresponding one of the plurality of lookup tables is selected.

As described above, the features of the present invention are summarized as follows.

(1) Bayer compression of low complexity

By applying LOCO-I modified Bayer image obtained through Bayer CFA to Bayer environment, the first predictor 211 or the first predictor 211 is used to reduce the computational complexity and use the G channel information closest to the resolution when compressing the R / B channel. The predictor 221 may estimate an appropriate previous value with a minimum of memory resources. In addition, there is an advantage that it is possible to save the computational resources by compressing immediately without any conversion process (Bayer RGB, Bayer YC). The invention of the low complexity Bayer compression technology developed through the present invention can reduce the memory resources by adding a little logic to the existing ISP, thereby lowering the ISP production cost.

2 scenes Adaptive  Bayer Compression

The first table generator 212 analyzes the difference histogram of the previous image to generate a table suitable for the current image. Through this, it is possible to adapt adaptively according to the scene, thereby obtaining better image quality than the fixed table method. In addition, the amount of change of the previous pixel obtained by the first predictor 211 or the second predictor 221 is checked by the first table selector 213 or the second table selector 222 to distinguish the edge region from the flat region, and to appropriately look up. Determine the table so that it is compressed more effectively than if you applied the same table. When measuring the change amount of the previous pixel, the information of the previous line is also used. In this case, the change amount is estimated by considering only the G channel which is most related to the resolution. This effectively distinguishes edge and flat areas, and reduces line memory to half because only G channels are considered.

(3) First  Encoder 216 and 2nd  Decoder (223)

In the present invention, a lookup table-based fixed length coding technique is used instead of an entropy coding technique used in JPEG, MPEG, and the like. The fixed length coding technique has the advantage of maintaining the same compression ratio for all scenes. Therefore, the complexity of the memory controller managing this is significantly lower than that of applying the entropy coding technique. This fixed length coding technique enables stable compression with less logic.

100: ISP system
110: lens shading unit 120: defect correction unit
130: WDR section 140: DNR section
150: demosaicing unit 160: color toning unit
170: memory 180: memory controller
210: encoding device 220: decoding device
211: first predictor 212: first table generator
213: first table selector 214: first difference value output
215: first adder 216: first encoder
221: second predictor 222: second table selector
223: second decoder 224: second adder

Claims (22)

An encoding method for a Bayer image including an RGB channel,
(a-1) estimating a value of a previous pixel using a plurality of pixels before the current pixel;
(b-1) estimating the amount of change between the plurality of pixels before the current pixel;
(c-1) generating a plurality of lookup tables using previous image frames;
(d-1) selecting one of a plurality of lookup tables generated in the step (c-1) using the change amount estimated in the step (b-1);
(e-1) differentially outputting the previous pixel estimated in the step (a-1) from the current pixel; And
(f-1) using the lookup table selected in step (d-1), receiving the difference value of step (e-1) and performing fixed length encoding; Characterized by the encoding method. delete The method of claim 1,
Step (a-1),
At least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And
And estimating a value of a previous pixel using at least one pixel belonging to a G channel of a previous line of the line to which the current pixel belongs. The method of claim 3,
In the step (a-1), when the current pixel G ₀ is a G channel,
A G ₁ pixel, which is a pixel of a G channel located immediately before the current pixel G ₀ , among pixels of a G channel;
The current pixel (G ₀₎ of a pixel in the G channel of the previous line of the line, the current pixel (G ₀₎ and the two adjacent pixels of the pixel of the G channel from the G ₂ pixel located behind;
The current pixel is (G ₀₎ of pixels of the G-channel which is located in front of the pixels of the G channel of the previous line of the line is located, the current pixel and the two pixels adjacent to (G _{0) 3} G pixel; And
Estimating a value of a previous pixel using a G ₄ pixel, which is a pixel of a G channel located in front of two pixels adjacent to the G ₁ pixel, among the pixels of the G channel of the previous line of the line where the G ₁ pixel is located; The encoding method characterized in that. The method of claim 3,
In the step (a-1), when the current pixel A ₀ is an R channel or a B channel,
An A ₁ pixel which is a pixel of the channel located immediately before the current pixel A ₀ among the pixels of the channel;
A G ₃ pixel that is a pixel of a G channel adjacent to the current pixel A ₀ , among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; And
An encoding method of estimating a value of a previous pixel using a G ₄ pixel which is a pixel of a G channel adjacent to the A ₁ pixel among the pixels of the G channel of the previous line of the line where the A ₁ pixel is located; . The method of claim 1,
Step (b-1),
At least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And
And a change amount is estimated using at least one pixel belonging to a G channel of a previous line of the line to which the current pixel belongs. The method of claim 6,
In the step (b-1), when the current pixel G ₀ is a G channel,
A G ₁ pixel, which is a pixel of a G channel located immediately before the current pixel G ₀ , among pixels of a G channel;
The current pixel (G ₀₎ of a pixel in the G channel of the previous line of the line, the current pixel (G ₀₎ and the two adjacent pixels of the pixel of the G channel from the G ₂ pixel located behind;
The current pixel is (G ₀₎ of pixels of the G-channel which is located in front of the pixels of the G channel of the previous line of the line is located, the current pixel and the two pixels adjacent to (G _{0) 3} G pixel;
Wherein the G ₁ pixels among the pixels of the G channel of the previous line of the line is located, the pixels of the G-channel which is located in front of the two pixels adjacent to the pixel G ₁ G ₄ pixels;
A G ₅ pixel which is a pixel of the G channel located immediately before the G ₁ pixel among the pixels of the G channel;
A G ₆ pixel that is a pixel of a G channel located immediately after the G ₂ pixels among the pixels of a G channel; And
Estimating a change amount by using a G ₇ pixel which is a pixel of a G channel located in front of two pixels adjacent to the G ₅ pixel among the pixels of the G channel of the previous line of the line where the G ₅ pixels are located; Encoding method. The method of claim 6,
In the step (b-1), when the current pixel A ₀ is an R channel or a B channel,
An A ₁ pixel which is a pixel of the channel located immediately before the current pixel A ₀ among the pixels of the channel;
An A ₂ pixel which is a pixel of the channel located immediately before the A ₁ pixel among the pixels of the channel;
A G ₃ pixel that is a pixel of a G channel adjacent to the current pixel A ₀ , among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located;
A G ₄ pixel that is a pixel of a G channel adjacent to the A ₁ pixel among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located;
A G ₇ pixel that is a pixel of a G channel adjacent to the A ₂ pixel among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located;
A G ₂ pixel which is a pixel of the G channel located immediately after the G ₃ pixels among the pixels of the G channel; And
And a variation amount is estimated using a G ₆ pixel which is a pixel of a G channel located immediately after the G ₂ pixels among the pixels of a G channel. The method of claim 1,
Step (c-1),
And for each of the pixels constituting the previous frame, generating the plurality of lookup tables using the density of the distribution of the difference values with adjacent previous pixels located immediately before the estimated. The method of claim 9,
Step (c-1),
Codes are generated by the first equation when the codes of the plurality of lookup tables are less than or equal to a predetermined threshold. An encoding method comprising generating a value. The method of claim 1,
Step (d-1),
And determining whether at least one of the plurality of changes estimated in the step (b-1) is equal to or greater than a value of a predetermined position of one lookup table of the plurality of lookup tables. The method of claim 11,
Step (d-1),
Selecting at least one lookup table when at least one of the plurality of changes estimated in step (b-1) is equal to or greater than a value of a predetermined position of one lookup table of the plurality of lookup tables,
When at least one of the plurality of changes estimated in step (b-1) is less than a value of a predetermined position of one of the plurality of lookup tables, a lookup table other than the corresponding one of the plurality of lookup tables Encoding method, characterized in that for selecting. A decoding method for a Bayer image including an RGB channel,
(a-2) estimating a value of a previous pixel using a plurality of pixels before the current pixel;
(b-2) estimating the amount of change between the plurality of pixels before the current pixel; And
(c-2) selecting one of a plurality of previously stored lookup tables by using the change amount estimated in the step (b-2). The method of claim 13,
The decoding method,
(d-2) performing fixed length decoding of the input data using the lookup table selected in step (c-2); And
and (e-2) adding and outputting the previous pixel estimated in the step (a-2) from the current pixel. The method of claim 14,
Step (a-2),
At least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And
And estimating a value of a previous pixel by using at least one pixel belonging to a G channel of a previous line of the line to which the current pixel belongs. The method of claim 15,
In the step (a-2), when the current pixel G ₀ is a G channel,
A G ₁ pixel, which is a pixel of a G channel located immediately before the current pixel G ₀ , among pixels of a G channel;
The current pixel (G ₀₎ of a pixel in the G channel of the previous line of the line, the current pixel (G ₀₎ and the two adjacent pixels of the pixel of the G channel from the G ₂ pixel located behind;
The current pixel is (G ₀₎ of pixels of the G-channel which is located in front of the pixels of the G channel of the previous line of the line is located, the current pixel and the two pixels adjacent to (G _{0) 3} G pixel; And
Estimating a value of a previous pixel using a G ₄ pixel, which is a pixel of a G channel located in front of two pixels adjacent to the G ₁ pixel, among the pixels of the G channel of the previous line of the line where the G ₁ pixel is located; Decoding method characterized in that. The method of claim 15,
In the step (a-2), when the current pixel A ₀ is an R channel or a B channel,
An A ₁ pixel which is a pixel of the channel located immediately before the current pixel A ₀ among the pixels of the channel;
A G ₃ pixel that is a pixel of a G channel adjacent to the current pixel A ₀ , among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located; And
A G ₄ pixel which is a pixel of a G channel adjacent to the A ₁ pixel among the pixels of the G channel of the previous line of the line where the A ₁ pixel is located; . The method of claim 14,
Step (b-2),
At least one pixel belonging to the same channel as the current pixel of the line to which the current pixel belongs; And
And estimating a change amount using at least one pixel belonging to a G channel of a previous line of the line to which the current pixel belongs. The method of claim 18,
In the step (b-2), when the current pixel G ₀ is a G channel,
A G ₁ pixel, which is a pixel of a G channel located immediately before the current pixel G ₀ , among pixels of a G channel;
The current pixel (G ₀₎ of a pixel in the G channel of the previous line of the line, the current pixel (G ₀₎ and the two adjacent pixels of the pixel of the G channel from the G ₂ pixel located behind;
The current pixel is (G ₀₎ of pixels of the G-channel which is located in front of the pixels of the G channel of the previous line of the line is located, the current pixel and the two pixels adjacent to (G _{0) 3} G pixel;
Wherein the G ₁ pixels among the pixels of the G channel of the previous line of the line is located, the pixels of the G-channel which is located in front of the two pixels adjacent to the pixel G ₁ G ₄ pixels;
A G ₅ pixel which is a pixel of the G channel located immediately before the G ₁ pixel among the pixels of the G channel;
A G ₆ pixel that is a pixel of a G channel located immediately after the G ₂ pixels among the pixels of a G channel; And
Estimating a change amount by using a G ₇ pixel which is a pixel of a G channel located in front of two pixels adjacent to the G ₅ pixel among the pixels of the G channel of the previous line of the line where the G ₅ pixels are located; Decoding method. The method of claim 18,
In the step (b-2), when the current pixel A ₀ is an R channel or a B channel,
An A ₁ pixel which is a pixel of the channel located immediately before the current pixel A ₀ among the pixels of the channel;
An A ₂ pixel which is a pixel of the channel located immediately before the A ₁ pixel among the pixels of the channel;
A G ₃ pixel that is a pixel of a G channel adjacent to the current pixel A ₀ , among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located;
A G ₄ pixel that is a pixel of a G channel adjacent to the A ₁ pixel among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located;
A G ₇ pixel that is a pixel of a G channel adjacent to the A ₂ pixel among pixels of a G channel of a previous line of the line where the current pixel A ₀ is located;
A G ₂ pixel which is a pixel of the G channel located immediately after the G ₃ pixels among the pixels of the G channel; And
And a G ₆ pixel, which is a pixel of a G channel located immediately after the G ₂ pixels among the pixels of a G channel, to estimate the amount of change. The method of claim 14,
Step (c-2),
And determining whether at least one of the plurality of changes estimated in the step (b-2) is equal to or greater than a value of a predetermined position of one lookup table of the plurality of lookup tables. The method of claim 21,
Step (c-2),
Selecting at least one lookup table when at least one of the plurality of changes estimated in step (b-2) is equal to or greater than a value of a predetermined position of one lookup table of the plurality of lookup tables,
If at least one of the plurality of changes estimated in the step (b-2) is less than a value of a predetermined position of one of the plurality of lookup tables, a lookup table other than the corresponding one of the plurality of lookup tables Decoding method characterized in that for selecting.

Images