KR20170075974A - Apparatus and Method for Encoding and Decoding of Image - Google Patents

Abstract

The present invention relates to a method and apparatus for encoding and decoding an image based on CALIC.
To this end, an image compression encoding method according to the present invention includes predicting a video signal value of a target pixel using neighboring pixels of a current pixel to be coded included in an image, and calculating a predicted video signal value and a video signal value A first residual signal generating step of generating a first residual signal according to a difference between the first prediction signal and the second prediction signal, a context classification step of analyzing the image and classifying the context of the target pixel, A prediction step of predicting a video signal value of the target pixel from the neighboring pixel of the target pixel, a second residual signal generating step of generating a second residual signal according to a difference between the video signal value predicted in the prediction step and the video signal value of the target pixel A second residual signal generating step for generating a second residual signal by subtracting the second residual signal from the second residual signal, And may include an enrichment step.

Description

[0001] The present invention relates to an image encoding and decoding method,

The present invention relates to a method and apparatus for encoding and decoding an image. In particular, the present invention relates to a method and apparatus for encoding and decoding an image based on CALIC.

Various algorithms exist for compressing images. Image compression is a method of reducing the amount of information by performing inter-frame prediction or performing intra-frame prediction, compressing only the difference image between the predicted image and the original image, and extracting redundant information among the respective information And a method of reducing the redundancy in terms of the encoding technique for the information itself.

In such image compression, there is a lossy compression method for achieving a high compression ratio, and a lossless compression method for recovering all information about an original image at a relatively low compression rate, while decoding the original image. Lossless compression methods are used in areas where reliability must be assured, such as medical imaging, legal data or remote sensing. Especially, in the field where image quality, precision and reliability are required, the usability of lossless compression is increasing.

There are techniques such as JPEG-LS and CALIC as lossless compression methods. In particular, CALIC (Context-Based, Adaptive, and Lossless Image Coding) is an algorithm that performs lossless compression based on context. However, the conventional CALIC technique has a problem that the compression rate is significantly lower as the size of the image increases.

(Non-Patent Document 0001) Wu, Xiaolin, and Nasir Memon. "Context-based, adaptive, lossless image coding." Communications, IEEE Transactions on 45.4 (1997): 437-444.

Accordingly, the present invention solves the problems of the conventional CALIC technique and provides a CALIC-based image compression method with improved compression ratio.

According to an aspect of the present invention, there is provided an image compression encoding method for predicting a video signal value of a target pixel using neighboring pixels of a current pixel to be coded included in an image, A first residual signal generating step of generating a first residual signal according to a difference between a video signal value of the target pixel and a first residual signal; A context classification step of analyzing the image and classifying the context of the target pixel; A prediction step of predicting a video signal value of the target pixel from the neighboring pixel of the target pixel using a prediction function set according to the classified context; A second residual signal generation step of generating a second residual signal according to a difference between the video signal value predicted in the prediction step and the video signal value of the target pixel; And an entropy encoding step of entropy encoding the second residual signal.

The first residual signal generating step may calculate a gradient in the target pixel and weight the neighboring pixels using a prediction coefficient set according to the calculated gradient direction to predict a video signal value of the target pixel .

Here, the context classification step may calculate the residual energy of each pixel of the image, and classify the context of the pixel according to the calculated residual energy.

The context classification step may calculate the gradient in each pixel of the image, and may calculate the residual energy using the calculated gradient and the first residual signal value in a neighboring pixel of the pixel.

The context classification step may calculate a texture value of each pixel of the image, and classify the context of the pixel according to the residual energy and the calculated texture value.

The context classification step may calculate the texture value according to a difference between an image signal value of a neighboring pixel of each pixel of the image and a predicted image signal value of the pixel in the first residual signal generation step have.

Here, the prediction step may predict a video signal value of the target pixel from the neighboring pixels by using the prediction function in which prediction coefficients are set for each classified context.

Here, the prediction coefficient may be a value of a cost function calculated according to a difference between an output value of the prediction function according to neighboring pixels of the pixel and a video signal value of the pixel, for each of the pixels included in each context, Is a coefficient that is set to be < RTI ID = 0.0 >

Wherein the prediction coefficient is a coefficient calculated to optimize the cost function set based on the minimum mean square error technique.

The entropy encoding step may entropy encode the second residual signal for each context.

According to another aspect of the present invention, there is provided an image compression decoding method including: entropy decoding step of entropy decoding a coded residual signal; A context determining step of determining a context of a pixel to be decoded; A prediction step of predicting a video signal value of the decoding target pixel from the previously decoded neighboring pixels of the decoding target pixel using a prediction function set according to the determined context; And calculating a video signal value of the pixel to be decoded using the predicted video signal value and the decoded residual signal.

The entropy decoding step may entropy decode the encoded residual signal according to the determined context.

The context determination step may calculate the residual energy and the texture value of the decoding target pixel using the previously decoded neighboring pixels and determine the context of the decoding target pixel according to the calculated residual energy and the texture value have.

Here, the prediction function may be a preset prediction function according to the determined context.

Wherein the predictive coefficient is a coefficient of the prediction function preset for each context, and for each of the pixels included in each context, an output value of the prediction function according to neighboring pixels of the pixel and a video signal value of the pixel And the cost function calculated according to the difference is set to be optimized.

Wherein the prediction coefficient is a coefficient calculated to optimize the cost function set based on the minimum mean square error technique.

According to another aspect of the present invention, there is provided an image compression encoding apparatus for predicting a video signal value of a target pixel using neighboring pixels of a target pixel included in an image, A first residual signal generating unit for generating a first residual signal according to a difference between a value of the target pixel and a video signal value of the target pixel; A context classifier for classifying the context of the target pixel by analyzing the image; A predictor for predicting a video signal value of the target pixel from the neighboring pixels of the target pixel using a prediction function set according to the classified context; A second residual signal generator for generating a second residual signal according to a difference between the video signal value predicted by the predictor and the video signal value of the target pixel; And an entropy encoding unit for entropy encoding the second residual signal.

Here, the prediction unit predicts a video signal value of the target pixel from the neighboring pixels using the prediction function in which a predictive coefficient is set for each classified context.

According to another aspect of the present invention, there is provided an apparatus for decoding an image, comprising: an entropy decoding unit for entropy decoding a coded residual signal; A context determination unit for determining a context of a pixel to be decoded; A predictor for predicting a video signal value of the decoding target pixel from the decoded neighboring pixels of the decoding target pixel using a prediction function set according to the determined context; And a decoding target pixel obtaining unit for calculating a video signal value of the decoding target pixel by using the predicted video signal value and the decoded residual signal.

Here, the prediction function may be a preset prediction function according to the determined context.

According to the image encoding and decoding method of the present invention, it is possible to achieve a higher compression rate when compressing a high-resolution image in the CALIC method in fields where precision and reliability such as medical images and legal data must be guaranteed.

1 is a flowchart of an image compression encoding method according to an embodiment of the present invention.
2 is a flowchart illustrating a method of compressing and decoding an image according to another embodiment of the present invention.
3 is a block diagram of an image compression encoding apparatus according to an embodiment of the present invention.
FIG. 4 is a block diagram of an image compression / decoding apparatus according to another embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to designate the same or similar components throughout the drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. In addition, the preferred embodiments of the present invention will be described below, but it is needless to say that the technical idea of the present invention is not limited thereto and can be variously modified by those skilled in the art.

There is a lossy compression method that achieves a high compression ratio, and a lossless compression method that restores all the information of the original image in decoding, though it has a relatively low compression rate. Lossless compression methods are used in areas where reliability must be assured, such as medical imaging, legal data or remote sensing. As such lossless compression methods, techniques such as JPEG-LS and CALIC exist.

Among the lossless compression methods, CALIC (Context-Based, Adaptive, and Lossless Image Coding) is a context-based lossless compression algorithm. ' Communications, IEEE Transactions on 45.4 (1997): 437-444. " The detailed operation method of the CALIC image coding method is as follows. First, scan the pixels - for example, raster scan - predict the target pixel using neighboring pixels of the target pixel. Next, a residual or an error image is generated according to the difference between the target pixel and the predicted pixel value. Next, the context of each pixel is classified by analyzing the error image. Here the context is classified with two criteria, one is the energy component of the error and one is the texture component. For example, if there are N classifications based on the energy of the error, and M classifications based on the texture component, there may be M x N total contexts. Next, entropy coding of the signal value of the pixel of the error image is performed for each context.

However, the existing CALIC technique as described above has a problem that the compression rate is significantly lower as the size of the image increases.

Accordingly, the present invention provides a CALIC image compression method with improved compression ratio.

According to another aspect of the present invention, there is provided an image compression method for predicting a target pixel using neighboring pixels of a target pixel included in an image, The method includes generating an image, analyzing the image to classify the context of each pixel, predicting the target pixel using the neighboring pixels of the target pixel included in the classified image for each of the classified contexts, Generating a second residual image corresponding to a difference between the image signals of the target pixels and entropy-coding the pixels of the second residual image for each context to perform image compression. In addition, the image decoding method according to the present invention can decode the encoded image code through the above process. Preferably, the video decoding method, the encoding method, and the apparatus according to the present invention can operate on the CALIC basis. However, if necessary, the image decoding method, the encoding method, and the apparatus according to the present invention may be applied to a lossless compression encoding / decoding codec other than a CALIC, and may be applied to a lossy compression /

In addition, the video encoding method, the decoding method, and the apparatus proposed by the present invention are devices and methods that can be used when lossless compression is to be performed on an image to be compression-encoded / decoded. Therefore, the compression target image can be a general image requiring lossless compression as described above. That is, the image coding method and decoding method and apparatus proposed in the present invention can be applied to general images for lossless compression.

Hereinafter, a CALIC-based image decoding method and a coding method according to the present invention will be described in detail.

1 is a flowchart of an image compression encoding method according to an embodiment of the present invention.

The image compression encoding method according to an exemplary embodiment of the present invention includes a first residual signal generation step S100, a context classification step S200, a prediction step S300, a second residual signal generation step S400, Encoding step S500.

The first residual signal generation step S100 predicts the video signal value of the target pixel using the neighboring pixels of the current pixel to be coded included in the video and outputs the predicted video signal value and the video signal value of the target pixel And generates a first residual signal according to the difference.

Here, the first residual signal generation step S100 may include calculating a gradient in the target pixel, weighting the neighboring pixels using a predictive coefficient set according to the calculated direction of the gradient, Can be predicted. Here, the gradient in the target pixel can be calculated as the gradient in the vertical direction and the gradient in the horizontal direction of the target pixel. Here, as a method of calculating a gradient in the vertical or horizontal direction, a known method of calculating a gradient in an image can be used. Also, in order to calculate the gradient value of the target pixel in the decoding process, the gradient can be calculated using the difference value between neighboring pixels of the target pixel.

Here, in order to calculate the gradient in the vertical direction, a gradient in the vertical direction can be calculated using neighboring pixels adjacent in the vertical direction among the neighboring pixels of the target pixel. For example, the difference between the left adjacent pixel and the adjacent pixel in the upper direction of the left adjacent pixel and the difference between the adjacent pixel in the upper adjacent direction and the adjacent pixel in the upward direction adjacent to the left adjacent pixel, Can be calculated. Similarly, in order to calculate the gradient in the horizontal direction, the horizontal gradient can be calculated using neighboring pixels neighboring in the horizontal direction among the neighboring pixels of the target pixel.

In the first residual signal generation step S100, the gradient of the target pixel is calculated, the direction of the gradient is determined, and the pixels adjacent to the target pixel are weighted by using the prediction coefficient set according to the direction, The video signal value of the pixel can be predicted. That is, the image signal value of the target pixel is calculated as a weighted sum of neighboring pixels. In this case, the prediction coefficient applied to the weighted sum may be set differently according to the direction of the gradient. For example, the predicted value of the video signal of the target pixel can be calculated as shown in the following equation (1).

는 대상 화소의 영상 신호 값의 예측 값이다.Where w _d is a prediction coefficient set according to the gradient direction d, I _n is the neighboring pixels of the target pixel,

Is a predicted value of a video signal value of a target pixel.

Here, the direction of the gradient can be determined using a horizontal gradient and a vertical gradient, and the direction of the gradient can be determined, for example, by comparing the difference value between the horizontal gradient and the vertical gradient to a predetermined threshold value. Here, the direction of the gradient may include information on the edge direction in the target pixel, and may further include information on the intensity of the edge. Here, the predetermined threshold value is a value that can be set as needed.

Here, in predicting the video signal value of the target pixel using the neighboring pixels of the target pixel in the first residual signal generation step S100, the prediction method in the existing CALIC technique can be used. "Wu, Xiaolin, and Nasir Memon." Context-based, adaptive, lossless image coding. " Communications, IEEE Transactions on 45.4 (1997): 437-444. "Using Gradient Adjusted Prediction (GAP), the video signal of a target pixel can be predicted from neighboring pixels.

Next, the difference between the video signal value predicted for the target pixel and the video signal value of the actual target pixel is calculated as described above, and the first residual signal is generated accordingly. In this way, a first residual image can be generated according to the first residual signal generated for each pixel of the image.

Next, the context classification step S200 classifies the context of the target pixel by analyzing the image. That is, the context may be classified for each pixel of the image to be encoded. Here, the context is predefined as a plurality of classes and can be defined as described below.

The context classification step S200 may calculate the residual energy of each pixel of the image and classify the context of the pixel according to the calculated residual energy. Residual energy can be calculated considering the size of the residual signal. In order to be able to determine the context of a specific pixel in the decoding step, the residual energy of a specific pixel can be calculated using the size of a residual signal of a neighboring pixel adjacent to the specific pixel.

Here, the residual energy can be calculated in consideration of the size of the residual signal of the neighboring pixel and the calculated gradient value of the specific pixel. For this, the context classification step S200 calculates a gradient in each pixel of the image, and calculates the residual energy using the calculated gradient and the first residual signal value in a neighboring pixel of the pixel .

Here, the gradient may be calculated according to a difference value between neighboring pixels in the same manner as described in the first residual signal generating step S100.

The context classification step S200 may calculate the residual energy as shown in Equation (2).

Where E _R is the residual energy, a, b, c is a constant that is set in advance, d _h can be the size of a residual signal in a horizontal direction of the gradient, d _v is a vertical gradient, e _n is the neighboring pixel have.

(i-1, j)|와 같이 각 산출될 수 있다. 여기서 I(i, j)는 레지듀얼 에너지를 산출하고자 하는 화소이고,

(i, j)는 제1 레지듀얼 신호 생성 단계(S100)에서 산출된 I(i, j)의 영상 신호 예측 값이다. 여기서 i는 영상의 가로 축, j는 영상의 세로 축의 인덱스를 각 나타내는 기호이다. 그리고 인덱스의 원점은 예를 들어 영상의 좌측 상단이 될 수 있고, 원점에서 오른쪽으로 진행할수록 i가 증가하고, 아래쪽으로 진행할수록 j가 증가하도록 인덱스가 설정될 수 있다.If the pixel to be calculated, where the residual energy as I (i, j), d h = | I (i-1, j) - I (i-2, j) | + I (i, j-1) - I (i-1, j-1) + | I (i, j- 1) - I (i + 1, j-1) |, d v = | I (i-1, j) - I (i-1, j-1) | + I (i, j-1) - I (i, j-2)) | + | I (i + 1, j-1) - I (i + 1, j-2) |, e n = | I (i-1, j) -

(i-1, j) |, respectively. Here, I (i, j) is a pixel for which residual energy is to be calculated,

(i, j) is a predicted video signal value of I (i, j) calculated in the first residual signal generation step S100. Where i is the horizontal axis of the image and j is the vertical axis index of the image. The origin of the index can be, for example, the upper left corner of the image, and the index i can be set so that i increases as it goes from the origin to the right and j increases as it goes down.

The context classification step S200 may determine the energy context according to the magnitude of the residual energy value calculated as above. For this, predetermined boundary values are set, and the energy context can be determined based on the boundary values. For example, the energy context can be divided into eight.

In addition, the context classification step S200 may calculate a texture value for each pixel of the image, and classify the context of the pixel according to the residual energy and the calculated texture value. Here, the texture can be defined according to the distribution of the image signal values of the pixels in the local region set around each pixel. In the present invention, the texture can be defined according to the difference between the video signal values of neighboring pixels of each pixel and the predicted video signal of the first residual signal for each pixel. In accordance with the difference between the video signal value of the neighboring pixel of each pixel of the image and the video signal value predicted by the first residual signal generating step S100 for each pixel, The texture value can be calculated.

For example, the texture value may be calculated according to the following equations (3) to (4).

는 텍스쳐 값을 산출하고자 하는 대상 화소의 영상 신호 값의 예측 값으로 상기 제1 레지듀얼 신호 생성 단계(S100)에서 산출된 값이다.Here, the texture value is Tc, bk is each bit value of each element or bit string of the texture value, Ink is the video signal value of the kth neighbor pixel,

Is a predicted value of a video signal value of a target pixel for which a texture value is to be calculated, and is a value calculated in the first residual signal generating step S100.

The texture Tc calculated as above can be represented by an array having K + 1 elements, and can be represented by a bit string according to the above arrangement. For example, when K is 7, Tc can be a number of 8 bits. The texture context can be defined for each texture value Tc.

Here, the context classification step (S200) can classify the context by considering the energy context and the texture context classified as above. For example, if there are N energy contexts and M texture contexts, there may be M x N total contexts.

Here, the context classification step S200 may classify the context of each pixel of the image using a CALIC algorithm. Here, "Wu, Xiaolin, and Nasir Memon." Context-based, adaptive, lossless image coding. Communications, IEEE Transactions on 45.4 (1997): 437-444. &Quot;.

Next, the prediction step S300 predicts the video signal value of the target pixel from the neighboring pixels of the target pixel using a prediction function set according to the classified context. Here, the prediction step S300 can predict the video signal value of the target pixel from the neighboring pixels by using the prediction function in which the prediction coefficient is set for each classified context.

In other words, in the prediction step S300, a prediction function having different prediction coefficients is set for each context, and when a specific pixel is classified into a specific context, the specific pixel can be predicted using the prediction function of the context. At this time, in the method of predicting a specific pixel, a prediction value of a video signal of a specific pixel can be calculated by using a video signal of a neighboring pixel neighboring to a specific pixel as an input of a prediction function. Here, the prediction function may be a function of multiplying the video signal values of neighboring pixels by a prediction coefficient and weighting the prediction coefficients.

For example, in the prediction step S300, the predicted value of the video signal of the target pixel can be calculated by Equation (5).

는 대상 화소의 영상 신호 값의 예측 값이다.Where w _C is a prediction coefficient set according to the context, I _n is the neighboring pixels of the target pixel,

Is a predicted value of a video signal value of a target pixel.

For example, I (i-1, j), I (i-1, j-1), I (i, j-1) 1, j-1), prediction coefficients (w1, w2, w3, w4) to be multiplied to each neighboring pixel can be set. Where i is the horizontal axis of the image and j is the vertical axis index of the image. The origin of the index can be, for example, the upper left corner of the image, and the index i can be set so that i increases as it goes from the origin to the right and j increases as it goes down. Such a prediction coefficient can be set for each context. That is, for example, if there are N contexts, there may be N sets of prediction coefficients. For example, when there are N contexts, there exists a set of prediction coefficients W_1 = (w1_1, w2_1, w3_1, w4_1) corresponding to the first context in the above example and a set W_k = (w1_k, w2_k, w3_k, w4_k) exist and a set W_N = (w1_N, w2_N, w3_N, w4_N) of prediction coefficients corresponding to the Nth context may exist.

According to the present invention, prediction can be performed more accurately by using a prediction function having different prediction coefficients for each context, thereby reducing the energy of the second residual signal as described below.

In order to set a prediction coefficient for each context, a prediction coefficient for achieving an optimal prediction for each context with respect to the learning image through a learning process may be calculated in advance and stored in the form of a table or the like. In the prediction step S300, the prediction coefficients corresponding to the context of the target pixel are retrieved by referring to the prediction coefficients preset and stored for each context, and the prediction value of the video signal of the target pixel can be calculated using the prediction coefficient.

For this purpose, the context of each pixel of the learning image is classified, and a cost function is designed according to the difference between the weighted sum of the corresponding pixel and the neighboring pixels of the corresponding pixel, and a prediction coefficient The predictive coefficient for each context can be obtained. In other words, the prediction coefficient may be a cost function calculated according to a difference between an output value of the prediction function according to neighboring pixels of the pixel and a video signal value of the pixel, for pixels included in each context in the context, Can be a coefficient set to be optimized. For example, the prediction coefficient may be a coefficient calculated to optimize the cost function based on a Minimum Mean Square Error (MMSE).

For example, the cost function can be set as shown in Equation (6) below.

Where I _t is the training image, W _C is the context-dependent prediction coefficient, and I _tn is the neighboring pixel in the training image.

Next, the second residual signal generation step S400 generates a second residual signal corresponding to the difference between the video signal value predicted in the prediction step S300 and the video signal value of the target pixel. Since the video signal values of the respective pixels are predicted according to the context in the prediction step S300, the second residual signal can be classified according to the context.

Next, the entropy encoding step (S500) entropy-codes the second residual signal.

Here, the entropy encoding step (S500) may entropy-encode the second residual signal for each context. That is, the second residual signal of each pixel classified according to each context in the image can be entropy-encoded by context. Here, entropy coding can use various known entropy coding techniques such as Huffman coding and arithmetic coding.

2 is a flowchart illustrating a method of compressing and decoding an image according to another embodiment of the present invention.

The image compression / decoding method according to another embodiment of the present invention may include an entropy decoding step (S1000), a context determining step (S2000), a prediction step (S3000), and a decoding target pixel acquiring step (S4000). Here, the decoding method can perform decoding in the reverse order of the encoding method described with reference to FIG. Therefore, the overlapping portions will be briefly described.

The entropy decoding step (S1000) entropy decodes the encoded residual signal.

Here, the entropy decoding step S1000 may perform the entropy decoding of the encoded residual signal according to the determined context.

The context determination step S2000 determines the context of the pixel to be decoded.

In the context determination step S2000, residual energy and texture value of the current pixel to be decoded are calculated using the previously decoded neighboring pixels, and the context of the current decoded pixel is calculated according to the calculated residual energy and texture value. Can be determined. Here, the method of determining the context may be in accordance with the method detailed in the context classification step S200 described above.

The prediction step S3000 predicts the video signal value of the decoding target pixel from the previously decoded neighboring pixels of the decoding target pixel using the prediction function set according to the determined context. Here, the prediction step S3000 may follow the method detailed in the prediction step S300 in the encoding step.

Here, the prediction function may be a preset prediction function according to the determined context. Here, the prediction coefficient is a coefficient of the prediction function preset for each context. Here, the prediction coefficient may be such that the cost function calculated according to the difference between the output value of the prediction function and the image signal value of the pixel according to the neighboring pixels of the pixel is optimal for the pixels included in each context It can be a set coefficient. In this case, the prediction coefficient may be a coefficient calculated to optimize the cost function set based on the minimum mean square error technique.

In the decoding target pixel acquiring step S4000, the video signal value of the decoding target pixel is calculated using the predicted video signal value and the decoded residual signal. Here, the predicted video signal value and the decoded residual signal may be summed to calculate the video signal value of the decoding target pixel.

3 is a block diagram of an image compression encoding apparatus according to another embodiment of the present invention.

The apparatus includes a first residual signal generating unit 100, a context classifying unit 200, a predicting unit 300, a second residual signal generating unit 400, And an encoding unit 500. Here, the image compression encoding apparatus according to the present invention can operate in the same manner as the image compression encoding method according to the present invention described in detail with reference to FIG. The overlapping portions will be briefly described.

The first residual signal generator 100 predicts the video signal value of the target pixel using the neighboring pixels of the current pixel to be coded included in the video and outputs the predicted video signal value and the video signal value of the target pixel And generates a first residual signal according to the difference.

The context classification unit 200 analyzes the image and classifies the context of the target pixel.

The prediction unit 300 predicts the video signal value of the target pixel from the neighboring pixels of the target pixel by using a prediction function set according to the classified context.

The second residual signal generator 400 generates a second residual signal according to the difference between the video signal value predicted by the predictor and the video signal value of the target pixel.

The entropy encoding unit 500 entropy-codes the second residual signal.

Here, the prediction unit 300 may predict the video signal value of the target pixel from the neighboring pixels by using the prediction function in which the prediction coefficient is set for each classified context.

FIG. 4 is a block diagram of an image compression / decoding apparatus according to another embodiment of the present invention.

The image compression / decoding method according to another embodiment of the present invention may include an entropy decoding unit 1000, a context determining unit 2000, a predicting unit 3000, and a decoding target pixel obtaining unit 4000. Here, the image compression / decoding apparatus according to the present invention can operate in the same manner as the image compression / decoding method according to the present invention described with reference to FIG. The overlapping portions will be briefly described.

The entropy decoding unit 1000 entropy-decodes the encoded residual signal.

The context determination unit 2000 determines the context of a pixel to be decoded.

The predicting unit 3000 predicts the video signal value of the decoding target pixel from the previously decoded neighboring pixels of the decoding target pixel by using a prediction function set according to the determined context.

Here, the prediction function may be a function in which a prediction coefficient is preset according to the determined context.

The pixel-to-be-decoded acquisition unit 4000 uses the predicted video signal value and the decoded residual signal to calculate the video signal value of the pixel to be decoded.

According to the present invention, it is possible to achieve a higher compression ratio when a high-resolution image is compressed by the CALIC method in a field where precision and reliability such as medical images and legal data must be guaranteed.

It is to be understood that the present invention is not limited to these embodiments, and all elements constituting the embodiment of the present invention described above are described as being combined or operated in one operation. That is, within the scope of the present invention, all of the components may be selectively coupled to one or more of them.

In addition, although all of the components may be implemented as one independent hardware, some or all of the components may be selectively combined to perform a part or all of the functions in one or a plurality of hardware. As shown in FIG. In addition, such a computer program may be stored in a computer readable medium such as a USB memory, a CD disk, a flash memory, etc., and read and executed by a computer to implement an embodiment of the present invention. As the recording medium of the computer program, a magnetic recording medium, an optical recording medium, a carrier wave medium, and the like can be included.

Furthermore, all terms including technical or scientific terms have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined in the Detailed Description. Commonly used terms, such as predefined terms, should be interpreted to be consistent with the contextual meanings of the related art, and are not to be construed as ideal or overly formal, unless expressly defined to the contrary.

It will be apparent to those skilled in the art that various modifications, substitutions and substitutions are possible, without departing from the scope and spirit of the invention as disclosed in the accompanying claims. will be. Therefore, the embodiments disclosed in the present invention and the accompanying drawings are intended to illustrate and not to limit the technical spirit of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments and the accompanying drawings . The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention.

S100: a first residual signal generation step
S200: Context classification step
S300: prediction step
S400: second residual signal generation step
S500: entropy encoding step
S1000: entropy decoding step
S2000: Context determination step
S3000: Forecast step
S4000: Pixel to be decoded
100: a first residual signal generating unit
200: Context classification section
300: prediction unit
400: second residual signal generating unit
500: Entropy encoding unit
1000: Entropy decoding unit
2000:
3000: prediction unit
4000: decoding target pixel obtaining unit

Claims (20)

In the image compression coding method,
Predicts a video signal value of the target pixel using neighboring pixels of a current pixel to be coded included in the video and generates a first residual signal according to a difference between the predicted video signal value and a video signal value of the target pixel Generating a first residual signal;
A context classification step of analyzing the image and classifying the context of the target pixel;
A prediction step of predicting a video signal value of the target pixel from the neighboring pixel of the target pixel using a prediction function set according to the classified context;
A second residual signal generation step of generating a second residual signal according to a difference between the video signal value predicted in the prediction step and the video signal value of the target pixel; And
And an entropy encoding step of entropy encoding the second residual signal. The method according to claim 1,
Wherein the first residual signal generating step calculates a gradient in the target pixel and weights the neighboring pixels using the prediction coefficient set according to the calculated gradient direction to predict the video signal value of the target pixel Wherein the image compression encoding method comprises: The method according to claim 1,
Wherein the context classification step calculates the residual energy of each pixel of the image and classifies the context of the pixel according to the calculated residual energy. The method of claim 3,
Wherein the context classification step calculates a gradient in each pixel of the image and calculates the residual energy using the calculated gradient and the first residual signal value in a neighboring pixel of the pixel Image compression encoding method. The method of claim 3,
Wherein the context classification step calculates a texture value for each pixel of the image and classifies the context of the pixel according to the residual energy and the calculated texture value. 6. The method of claim 5,
The context classification step calculates the texture value according to a difference between a video signal value of a neighboring pixel of each pixel of the image and a video signal value predicted in the first residual signal generation step for each pixel To be encoded. The method according to claim 1,
Wherein the predicting step predicting a video signal value of the target pixel from the neighboring pixels using the prediction function in which a predictive coefficient is set for each of the classified contexts. 8. The method of claim 7,
Wherein the predictive coefficient is calculated such that a cost function calculated according to a difference between an output value of the prediction function according to neighboring pixels of the pixel and a video signal value of the pixel is optimized for pixels included in each context in each context Wherein the coefficient is a set coefficient. 9. The method of claim 8,
Wherein the prediction coefficient is a coefficient calculated to optimize the cost function based on a minimum mean square error technique. The method according to claim 1,
Wherein the entropy coding step entropy-codes the second residual signal for each context. An image compression decoding method comprising:
An entropy decoding step of entropy decoding the encoded residual signal;
A context determining step of determining a context of a pixel to be decoded;
A prediction step of predicting a video signal value of the decoding target pixel from the previously decoded neighboring pixels of the decoding target pixel using a prediction function set according to the determined context;
And calculating a video signal value of the pixel to be decoded using the predicted video signal value and the decoded residual signal. 12. The method of claim 11,
Wherein the entropy decoding step entropy decodes the encoded residual signal according to the determined context. 12. The method of claim 11,
The context determination step may include calculating the residual energy and the texture value of the decoding target pixel using the previously decoded neighboring pixels and determining the context of the decoding target pixel according to the calculated residual energy and the texture value Wherein the image compression / 12. The method of claim 11,
Wherein the prediction function is a preset predictive function according to the determined context. 15. The apparatus of claim 14,
A coefficient of the prediction function preset for each context,
And a cost function calculated according to a difference between an output value of the prediction function according to neighboring pixels of the pixel and a video signal value of the pixel with respect to the pixels included in each of the contexts, / RTI > 16. The method of claim 15,
Wherein the prediction coefficient is a coefficient calculated to optimize the cost function based on a minimum mean square error technique. An image compression encoding apparatus comprising:
Predicts a video signal value of the target pixel using neighboring pixels of a current pixel to be coded included in the video and generates a first residual signal according to a difference between the predicted video signal value and a video signal value of the target pixel A first residual signal generator;
A context classifier for classifying the context of the target pixel by analyzing the image;
A predictor for predicting a video signal value of the target pixel from the neighboring pixels of the target pixel using a prediction function set according to the classified context;
A second residual signal generator for generating a second residual signal according to a difference between the video signal value predicted by the predictor and the video signal value of the target pixel; And
And an entropy coding unit for entropy-coding the second residual signal. 18. The method of claim 17,
Wherein the prediction unit predicts a video signal value of the target pixel from the neighboring pixels using the prediction function in which prediction coefficients are set for each classified context. An image compression / decoding apparatus comprising:
An entropy decoding unit for entropy decoding the encoded residual signal;
A context determination unit for determining a context of a pixel to be decoded;
A predictor for predicting a video signal value of the decoding target pixel from the decoded neighboring pixels of the decoding target pixel using a prediction function set according to the determined context;
And a decoding target pixel obtaining unit for calculating a video signal value of the decoding target pixel by using the predicted video signal value and the decoded residual signal. 20. The method of claim 19,
Wherein the prediction function is a preset predictive function according to the determined context.

Images