WO2012011670A2

WO2012011670A2 - Method and device for predicting color space using plurality of frequency domain weighting prediction filters, and method and device for encoding/decoding image using same

Info

Publication number: WO2012011670A2
Application number: PCT/KR2011/004513
Authority: WO
Inventors: 송진한; 임정연; 최윤식; 김용구; 최영호; 정진우
Original assignee: 에스케이텔레콤 주식회사
Priority date: 2010-07-23
Filing date: 2011-06-21
Publication date: 2012-01-26
Also published as: WO2012011670A3; KR20120011353A; KR101681307B1

Abstract

One embodiment of the present invention relates to a method and a device for predicting a color space using a plurality of frequency domain weighting prediction filters, and to a method and a device for encoding/decoding an image using same. According to one embodiment, the present invention provides the device and the method for predicting the color space comprising: a prediction weight calculating unit for determining a weighted block from the encoded/decoded peripheral blocks of the current block, and for calculating the prediction weight for each of the frequency domains using the weighted block; a prediction frequency selection unit for calculating a prediction gain for each of the frequency domains from the prediction weight for each of the frequency domains, and for selecting the frequency domain to be used for predicting the color space by using the prediction gain for each of the frequency domains; and a frequency domain prediction unit for receiving remaining transformed blocks, and for predicting the color space in the remaining transformed blocks from the prediction gain of the selected frequency domain. The present invention also provides the device and the method for predicting the color space and the method and the device for encoding/decoding the image using same, comprising: the prediction weight calculating unit for determining the weighted block from the encoded/decoded peripheral blocks of the current block, and for calculating the prediction weight for each of the frequency domains using the current weighted block; the prediction frequency selection unit for calculating the prediction gain for each of the frequency domains from the prediction weight for each of the frequency domains, and for selecting the frequency domain to be used for predicting the color space by using the prediction gain for each of the frequency domains; and a frequency domain recovery unit for recovering the remaining transformed blocks from the color space prediction block using the prediction gain of the selected frequency domain.

Description

Color space prediction method and apparatus using frequency domain multiple weighted prediction filter and image encoding / decoding method and apparatus using same

An embodiment of the present invention relates to a color space prediction method and apparatus using a frequency domain multiple weighted prediction filter, and an image encoding / decoding method and apparatus using the same. More specifically, in the frequency domain prediction of the color-space residual signal, in order to efficiently use the color component correlation that may exist differently in the unit image, the RGB color-space is optimized by using the optimal one among the plurality of prediction filters for each unit image. The present invention relates to a color space prediction method and apparatus using a frequency domain multiple weighted prediction filter for improving the compression performance of video data, and to an image encoding / decoding method and apparatus using the same.

Recently, as the demand for the application of ultra-high definition video such as digital cinema, medical image, and digital archive increases, technologies for directly encoding / decoding in the RGB color space have been developed. This has a problem such that a rounding error occurs when the encoding / decoding in the widely used YCbCr color space is converted to the RGB color space. Therefore, H.264 / AVC, the latest video compression standard, supports encoding / decoding in RGB color space.

In the RGB color space, there is often a high similarity between color planes. However, H.264 / AVC does not include a technique for improving compression efficiency by directly using the similarity of RGB color spaces. Recently, techniques for improving coding efficiency by performing direct prediction in the RGB color space have been reported. These techniques can be broadly classified into techniques that predict in the pixel domain and those that predict in the frequency domain. The techniques predicted in the frequency domain provide better compression performance than those predicted in the pixel domain. It is known to provide.

The frequency domain prediction technique converts an original signal into a frequency domain by performing a frequency transform such as a discrete cosine transform (DCT), and then performs prediction between RGB color components in the frequency domain. This is based on the fact that there is a high correlation in the frequency domain between RGB color components. In addition, when the prediction is performed in the frequency domain, since the noise is mostly in the high frequency region, there is an advantage that the noise is prevented from propagating to other color components by predicting only a specific region. However, the conventional frequency domain prediction technique has a problem in that it does not properly use the similarity between color components that may exist differently for each image by using the same prediction function for all images.

In order to solve this problem, there is a technique of adaptively determining the inter-color prediction filter according to the characteristics of the recent image. In this technique, the prediction filter is adaptively determined for each image unit such as a macroblock or a basic unit, thereby improving the compression performance of the existing frequency domain prediction technique. In this technique, when the adaptively determined image unit is too small, it is difficult to secure statistical stability in calculating the prediction filter of the image unit to be encoded. In order to prevent this, when the image unit is enlarged, there is a possibility of having various image characteristics in one image unit. Therefore, there is a problem in that the prediction efficiency may decrease because only one prediction filter does not properly reflect the color plane correlations of all blocks in the image unit.

In order to solve this problem, an embodiment of the present invention uses a plurality of prediction filters for each unit image to efficiently use color component correlation that may exist differently in the unit image when frequency-domain prediction of the color-space residual signal is performed. The main purpose is to further increase the encoding / decoding efficiency.

In order to achieve the above object, according to an embodiment of the present invention, in an apparatus for encoding / decoding an image, a prediction block is generated by predicting a current block for each color plane, and the prediction block is subtracted from the current block. Generating a block, transforming the residual block, and applying the transform coefficients of the blocks in the weighted candidate block set to each of the N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. Selecting the prediction weights and candidate frequency domains for each frequency domain to be used for the color space prediction based on the frequency domains having the optimal prediction efficiency among the frequency weighted prediction weights and candidate frequency domains calculated for each of the N weighted candidate blocksets. Color for the transformed residual block using prediction weights and candidate frequency domain Video encoder for encoding the color space prediction block generating the inter prediction block, and; And decoding the color space prediction block by decoding the encoded data, and transforming the block within the weighted candidate block set for each of the N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. A frequency domain prediction weight and a candidate frequency domain to be used for color space prediction are selected based on coefficients, and a frequency having an optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weight candidate blocksets. Reconstruct a residual block transformed from the color space prediction block using a prediction weight for each region and a candidate frequency domain, generate a prediction block by predicting a current block, and add the reconstructed residual block and the prediction block to the current block. Comprising an image decoder for reconstructing the block It provides a video encoding / decoding apparatus as set.

According to another aspect of the present invention, there is provided an apparatus for encoding an image, comprising: a predictor configured to generate a prediction block by predicting a current block for each color plane; A subtractor for generating a residual block by subtracting the prediction block from the current block; A transformer for transforming the residual block; Prediction weights for each frequency domain to be used for color space prediction based on the transform coefficients of blocks in the weighted candidate block set for each of the weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. And selecting the candidate frequency domain and converting the prediction weighted frequency and the candidate frequency domain by the frequency domain with the optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weighted candidate blocksets. A color space predictor for generating a color space prediction block for the residual block; And an encoder for encoding the color space prediction block.

According to another aspect of the present invention, there is provided an apparatus for decoding an image, the apparatus comprising: a decoder for decoding color space prediction blocks by decoding encoded data; Prediction weights for each frequency domain to be used for color space prediction based on the transform coefficients of blocks in the weighted candidate block set for each of the weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. And selecting the candidate frequency domain and using the prediction domain weights and the candidate frequency domains having the optimal prediction efficiency among the frequency domain prediction weights and the candidate frequency domains calculated for each of the N weight candidate blocksets. A color space reconstructor for reconstructing the residual block transformed from the predictive block; An inverse transformer for inversely transforming the transformed residual block to restore the residual block; A predictor for predicting a current block to generate a predicted block; And an adder for reconstructing the current block by adding the reconstructed residual block and the prediction block.

In addition, to achieve another object of the present invention, an embodiment of the present invention, in the color space prediction apparatus of the transformed residual block, N (N is a natural number of two or more) from the transform coefficient block of the decoded residual signal A weighted candidate selection unit for determining a weighted candidate block set; For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated using a transform coefficient of a block in the weighted candidate block set, and a prediction gain for each frequency domain is calculated using the prediction weights for each frequency domain. A candidate weight frequency calculator for selecting a candidate frequency domain to be used for color space prediction by using the prediction gain for each frequency domain; An optimum weighting filter determiner for receiving prediction weights and N candidate frequency domains for each of the N frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights and the candidate frequency domains for each frequency domain having an optimal prediction efficiency; And a frequency domain predictor configured to generate a color space prediction block for the transformed residual block by using the prediction weights for each frequency domain and the candidate frequency domain having the optimal prediction efficiency. do.

Here, the N weighted candidate blocksets may be a set of neighboring blocks and a set of M-blocked macroblocks (M is a natural number of 1 or more).

Herein, the N weight candidate blocks may be a set of macroblocks for each filter type of a macroblock belonging to a decoded macroblock row.

The color space prediction apparatus may include: a row prediction detector configured to determine whether color space prediction of one macroblock row is completed; And after the color space prediction of the one macroblock row is completed, the weight of each type, which is a prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of macroblocks for each filter type of the one macroblock row. The type-weighting value may further include a filter type adjusting unit for resetting the filter type for the macroblock according to the type-specific prediction weighting value having the optimal filter efficiency for all the macroblocks in the single macroblock row.

The optimal filter rate may be a case where the rate-distortion cost is the smallest in consideration of the amount of bits and distortions generated when the current block is predicted and encoded for each type-weight.

The filter type adjusting unit may repeat the resetting of the filter type when the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.

The prediction weight for each frequency domain may be calculated by a correlation between the reference color plane and the remaining color planes.

The prediction weights and the candidate frequency domains for each frequency domain having the optimal prediction efficiency are generated when the current block is predicted and encoded for each of the frequency weighted prediction weights and candidate frequency domains transmitted from the candidate weighted frequency calculator. In consideration of the amount of bits and the distortion value, the prediction weight and the candidate frequency domain for each frequency domain when the rate-distortion cost is the smallest can be included.

The frequency domain prediction unit subtracts a value obtained by applying a prediction weight of the selected frequency domain to frequency coefficients of color planes other than the reference color plane from frequency coefficients of the selected frequency domain of the reference color plane of the transformed residual block. Color space prediction can be performed.

The frequency domain predictor may perform color space prediction after quantizing the prediction weights of the selected frequency domain.

The prediction weight for each frequency domain may be calculated in any one of a sequence unit, a frame unit, a macroblock unit, and a subblock unit of an image.

Further, to achieve another object of the present invention, an embodiment of the present invention, in the color space prediction device of the color space prediction block, N (N is a natural number of two or more) from the transform coefficient block of the decoded residual signal A weighted candidate selection unit for determining a weighted candidate block set; For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated using a transform coefficient of a block in the weighted candidate block set, and a prediction gain for each frequency domain is calculated using the prediction weights for each frequency domain. A candidate weight frequency calculator for selecting a candidate frequency domain to be used for color space prediction by using the prediction gain for each frequency domain; An optimum weighting filter determiner for receiving prediction weights and N candidate frequency domains for each of the N frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights and the candidate frequency domains for each frequency domain having an optimal prediction efficiency; And a frequency domain reconstruction unit for reconstructing the residual block converted from the color space prediction block by using the prediction weight for each frequency domain and the candidate frequency domain having the optimal prediction efficiency. .

Here, the color space prediction apparatus may determine whether color space prediction of one macroblock row is completed.

Here, after the color space prediction of the one macroblock row is completed, the weight of each type, which is the prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of the macroblocks for each filter type of the one macroblock row. The type-weighting value may further include a filter type adjusting unit for resetting the filter type for the macroblock according to the type-specific prediction weighting value having the optimum filter efficiency for all the macroblocks in the one macroblock row.

Here, the optimal filter rate may be a case where the rate-distortion cost is the smallest in consideration of the bit amount and the distortion value generated when the current block is predicted and encoded with respect to each type weight.

Here, the filter type adjusting unit may repeat the resetting of the filter type when the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.

In this case, the prediction weights and the candidate frequency domains for each frequency domain having the optimal prediction efficiency are used for predicting and encoding the current block for each of the prediction weights and the candidate frequency domains for each frequency domain transmitted from the candidate weight frequency calculation unit. It may be a prediction weight for each frequency domain and a candidate frequency domain when the rate-distortion cost is the smallest in consideration of the amount of bits and distortions that occur.

Here, the frequency domain reconstruction unit adds a value obtained by applying a prediction weight of the selected frequency domain to a frequency coefficient of the selected frequency domain of the reference color plane of the color space prediction block and a frequency coefficient of color planes other than the reference color plane. By doing so, the converted residual block can be restored.

The frequency domain reconstructor may perform color space prediction after quantizing the prediction weights of the selected frequency domain.

Here, the prediction weight for each frequency domain may be calculated in any one of a sequence unit, a frame unit, a macroblock unit, and a subblock unit of an image.

In addition, to achieve another object of the present invention, an embodiment of the present invention, in a method of encoding / decoding an image, predicts a current block for each color plane to generate a prediction block, the prediction block in the current block Subtract a to generate a residual block, transform the residual block, and block N in the weighted candidate block set for each of the N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. The prediction weights and candidate frequency domains of each frequency domain to be used for the color space prediction are selected based on the transform coefficient of and the optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weighted candidate blocksets is obtained. The transformed residual using the prediction weights for each frequency domain and the candidate frequency domain Generating a color space prediction block for the green, and the step of encoding the color space prediction block; And decoding the color space prediction block by decoding the encoded data, and transforming the block within the weighted candidate block set for each of the N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. A frequency domain prediction weight and a candidate frequency domain to be used for color space prediction are selected based on coefficients, and a frequency having an optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weight candidate blocksets. Reconstruct the residual block transformed from the color space prediction block by using the prediction weights and the candidate frequency domain for each region, inversely transform the transformed residual block, reconstruct the residual block, predict the current block, and generate a prediction block, The reconstructed residual block is added to the prediction block to add an image. It provides a video encoding / decoding method comprising the step of reconstructing the current block.

In addition, to achieve another object of the present invention, an embodiment of the present invention, a method for encoding an image, comprising: generating a prediction block by predicting the current block for each color plane; Subtracting the prediction block from the current block to generate a residual block; Transforming the residual block; Prediction weights for each frequency domain to be used for color space prediction based on the transform coefficients of blocks in the weighted candidate block set for each of the weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. And selecting the candidate frequency domain and converting the prediction weighted frequency and the candidate frequency domain by the frequency domain with the optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weighted candidate blocksets. A color space prediction step of generating a color space prediction block for the residual block; And encoding the color space prediction block.

According to another aspect of the present invention, there is provided a method of decoding an image, the method comprising: decoding a color space prediction block by decoding encoded data; Prediction weights for each frequency domain to be used for color space prediction based on the transform coefficients of blocks in the weighted candidate block set for each of the weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. And selecting the candidate frequency domain and using the prediction domain weights and the candidate frequency domains having the optimal prediction efficiency among the frequency domain prediction weights and the candidate frequency domains calculated for each of the N weight candidate blocksets. A color space reconstruction step of reconstructing the residual block transformed from the prediction block; An inverse transform step of inversely transforming the transformed residual block to restore a residual block; Predicting a current block to generate a predicted block; And reconstructing the current block by adding the reconstructed residual block and the prediction block.

In addition, to achieve another object of the present invention, an embodiment of the present invention, in the color space prediction method of the transformed residual block, N (N is a natural number of 2 or more) from the transform coefficient block of the decoded residual signal Determining a weighted candidate block set; For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated based on the transform coefficients of the blocks in the weighting candidate block set, a prediction gain for each frequency domain is calculated from the prediction weights for the frequency domain, and the frequency domain is calculated. Selecting a candidate frequency region to be used for color space prediction by using the predictive gain for each star; Receiving prediction weights for each N frequency domains and N candidate frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights for each frequency domain and the candidate frequency domains having optimal prediction efficiency; And a frequency domain prediction step of generating a color space prediction block for the transformed residual block using the prediction domain weight and the candidate frequency domain for each frequency domain having the optimal prediction efficiency. to provide.

Further, to achieve another object of the present invention, an embodiment of the present invention, in the color space prediction method of the color space prediction block, N (N is a natural number of two or more) from the transform coefficient block of the decoded residual signal Determining a weighted candidate block set; For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated based on the transform coefficients of the blocks in the weighting candidate block set, a prediction gain for each frequency domain is calculated from the prediction weights for the frequency domain, and the frequency domain is calculated. Selecting a candidate frequency region to be used for color space prediction by using the predictive gain for each star; Receiving prediction weights for each N frequency domains and N candidate frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights for each frequency domain and the candidate frequency domains having optimal prediction efficiency; And a frequency domain reconstruction step of reconstructing the residual block transformed from the color space prediction block by using the prediction domain weight and the candidate frequency domain for each frequency domain having the optimal prediction efficiency. do.

As described above, according to the exemplary embodiment of the present invention, when predicting the frequency domain of the color-space residual signal, a plurality of predictions are performed for each unit image in order to efficiently use color component correlation that may exist differently in the unit image. By using an optimal filter, the encoding / decoding performance of RGB color-space video data can be improved.

1 is a block diagram illustrating a color space prediction apparatus 100 according to a first embodiment of the present invention.

2 is a diagram illustrating a set of neighboring blocks of a current block.

3 is a diagram illustrating a decoded M macroblock row set.

FIG. 4 is a diagram illustrating a set of macroblocks by weight filter type of a macroblock belonging to an upper macroblock row of a current block.

5 is a diagram illustrating color space prediction performed by a color space prediction apparatus according to a first embodiment of the present invention with respect to an RGB input image.

FIG. 6 illustrates block format and color space prediction for an RGB image when 4x4 DCT is performed on one macroblock when the prediction reference color plane is a G plane.

7 is a diagram illustrating before (a) and after (b) resetting the filter type.

8 is a diagram illustrating a color space prediction apparatus 800 according to a second embodiment of the present invention.

9 is a block diagram schematically illustrating a video encoding apparatus according to an embodiment of the present invention.

10 is a block diagram schematically illustrating a configuration of an image decoding apparatus according to an embodiment of the present invention.

11 is a flowchart illustrating a color space prediction method according to a first embodiment of the present invention.

12 is a flowchart illustrating a color space prediction method according to a second embodiment of the present invention.

13 is a flowchart illustrating a video encoding method according to an embodiment of the present invention.

14 is a flowchart illustrating an image decoding method according to an embodiment of the present invention.

Hereinafter, some embodiments of the present invention will be described in detail through exemplary drawings. In adding reference numerals to the components of each drawing, it should be noted that the same reference numerals are assigned to the same components as much as possible even though they are shown in different drawings. In addition, in describing the present invention, when it is determined that the detailed description of the related well-known configuration or function may obscure the gist of the present invention, the detailed description thereof will be omitted.

In addition, in describing the component of this invention, terms, such as 1st, 2nd, A, B, (a), (b), can be used. These terms are only for distinguishing the components from other components, and the nature, order or order of the components are not limited by the terms. If a component is described as being "connected", "coupled" or "connected" to another component, that component may be directly connected to or connected to that other component, but there may be another configuration between each component. It is to be understood that the elements may be "connected", "coupled" or "connected".

A video encoding apparatus (Video Encoding Apparatus), a video decoding apparatus (Video Decoding Apparatus), and a color space prediction apparatus to be described below may be a personal computer (PC), a notebook computer, a personal digital assistant (PDA), A user terminal such as a portable multimedia player (PMP), a PlayStation Portable (PSP), a wireless communication terminal, a smart phone, or a server terminal such as an application server or a service server. A communication device such as a communication modem for communicating with various devices or a wired / wireless communication network, a memory for storing various programs and data for encoding or decoding an image or inter or intra prediction for encoding or decoding Microprocessors for Execution and Operation and Control It may refer to a different apparatus having a.

In addition, the image encoded in the bitstream by the video encoding apparatus is real-time or non-real-time through the wired or wireless communication network, such as the Internet, local area wireless communication network, wireless LAN network, WiBro network, mobile communication network, or the like, or a cable, universal serial bus (USB: Universal) It may be transmitted to an image decoding apparatus through various communication interfaces such as a serial bus, and may be decoded by the image decoding apparatus to restore and reproduce the image.

In general, a video may be composed of a series of pictures, and each picture may be divided into a predetermined area such as a frame or a block. When a region of an image is divided into blocks, the divided blocks may be classified into intra blocks and inter blocks according to an encoding method. An intra block refers to a block that is encoded by using an intra prediction coding scheme. Intra prediction coding is performed by using pixels of blocks that have been previously encoded, decoded, and reconstructed in a current picture that performs current encoding. A prediction block is generated by predicting pixels of a block, and a difference value with pixels of the current block is encoded. An inter block refers to a block that is encoded using inter prediction coding. Inter prediction coding generates a prediction block by predicting a current block in a current picture by referring to one or more past pictures or future pictures, and then generates a current block. This is a method of encoding the difference value with. Here, a frame referred to for encoding or decoding the current picture is referred to as a reference frame.

The color space prediction apparatus 100 according to the first exemplary embodiment of the present invention includes a weight candidate candidate 110, a candidate weight frequency calculator 120, an optimal weighted filter determiner 130, and a frequency domain predictor 140. In some cases, a conversion unit 150, a row prediction detection unit 160, and a filter type adjustment unit 170 may be added.

The weight candidate selection unit 110 determines a set of N weight candidate blocks from the transform coefficient block of the decoded residual signal. Here, the "decoding" in "conversion coefficient block of decoded residual signal" means that decoding is performed for use as a reference block after the block is encoded in the image encoding apparatus.

The weighted candidate selecting unit 110 is a set of N weighted candidate block sets determined from transform coefficient blocks of a decoded residual signal, and a macro of M (M is one or more natural numbers) already decoded. It may be a block row set.

The candidate weighted frequency calculating unit 120 calculates a prediction weight for each frequency domain based on the transform coefficients of blocks in the weighted candidate block set for each of the N weighted candidate block sets, and predicts and obtains the prediction gain for each frequency domain from the prediction weights for the frequency domain. Next, the candidate frequency domain to be used for color space prediction is selected using the prediction gain for each frequency domain.

The optimal weighted filter determiner 130 receives the prediction weights for each N frequency domains and the N candidate frequency domains, calculates and compares the prediction efficiency for each, and compares the prediction weights for each frequency domain and the candidate frequency domains with the optimal prediction efficiency. Determine.

The frequency domain prediction unit 140 generates a color space prediction block for the residual block transformed by using the prediction weight for each frequency domain and the candidate frequency domain having an optimal prediction efficiency.

The transformer 150 receives an intra or inter predicted residual block and performs a transform such as DCT.

The row prediction detector 160 checks whether the color space prediction of one macroblock row is completed.

After the color space prediction of one macroblock row is completed, the filter type adjusting unit 170 performs weighting for each type, which is a prediction weight for each frequency domain, based on the transform coefficient of the macroblock for each set of macroblocks for each filter type of one macroblock row. Calculate and reset the filter type for the macroblock according to the type-specific prediction weights with the optimal filter efficiency for all macroblocks in one macroblock row.

FIG. 2 is a diagram illustrating a set of neighboring blocks of a current block, and FIG. 3 is a diagram illustrating a set of M decoded macroblock rows that are decoded.

The information on the set of neighboring blocks of the current block and the information on the decoded (or coded) M macroblock row sets as shown in FIGS. 2 and 3 respectively indicate the information on the first block set and the second block. Information on the set may be transmitted to the candidate weight frequency calculator 120.

The decoded macroblock is a macroblock existing above and to the left of the current macroblock since the decoded macroblock is decoded in a raster scan order. Since the statistical similarity with the current macroblock is high in the block close to the current macroblock, a prediction filter can be obtained using surrounding macroblocks as shown in FIG. 2. As shown in FIG. 2, the first block set may be determined as a pre-decoded left (MB_L), top (MB_U), top left (MB_UL), and top right (MB_UR) macroblocks of the current macroblock. As described above, when the prediction filter is calculated using only the neighboring macroblocks closest to the current macroblock, the statistical similarity with the current macroblock may be the highest and thus the prediction efficiency may be high. However, since the number of samples for calculating the predictive filter is small, it is difficult to secure statistical stability in the calculation of the predictive filter of the current image unit to be encoded. Therefore, in order to prevent this, the second block set may calculate the prediction filter by increasing the number of samples for obtaining the prediction filter.

As shown in FIG. 3, macroblocks existing in macroblock rows from k-1 to macroblocks of the k-th macroblock row pre-coded in order to calculate a prediction filter of the current macroblock located in the K-th macroblock row are second. Decide on a set of blocks. Where n can be an adaptive or fixed value. In this way, when the second block set is determined, the number of macroblocks belonging to the second block set is n * M, assuming that the number of macroblocks constituting one macroblock row is M. If n or M is large, the statistical stability of the prediction filter can be sufficiently secured, and the prediction filter needs to be calculated only once for each macroblock row. However, when the image unit for obtaining the prediction filter is enlarged, there may be a possibility of having various image characteristics in one image unit. Therefore, there is a limitation that the correlation with the current macroblock may be degraded, but the prediction efficiency may be reduced. Therefore, according to the present invention, it is possible to provide a higher prediction efficiency by adaptively determining, for each macroblock, a more efficient one of the prediction filter calculated from the first block set and the prediction filter calculated from the second block set.

FIG. 4 is a diagram illustrating a set of macroblocks by weight filter type of a macroblock belonging to an upper macroblock row of a current block. In this case, the weight filter may mean a weight for each frequency coefficient of the residual block that can be calculated for each group of the macroblocks MB_A and MB_B.

In relation to the decoded (or decoded) macroblock row as shown in FIG. 4, the first block set to the N-th block set information may be transmitted to the candidate weighted frequency calculating unit 120. The description is based on the assumption that two block set information related to one completed (or encoded) upper macroblock row is generated, but the number of decoded (or encoded) target macroblock rows and the number of generated block sets are implemented. It may vary depending on the example, and a person skilled in the art can easily extend it.

The candidate weight frequency calculating unit 120 calculates the prediction weights for each frequency domain based on the transform coefficients of the weighted candidate blocks for each block set, calculates the prediction gains for each frequency domain from the calculated prediction weights for the frequency domains, and calculates A candidate frequency domain to be used for color space prediction is selected using the prediction gain for each frequency domain.

Here, only the functions of the first predictive weight calculator 122 and the first predictive frequency selector 124 in the candidate weight frequency calculator 120 will be described. The functions of the N-th predictive weight calculating unit 126 and the N-th predictive frequency selecting unit 128 differ only in the block set, and the remaining operations are performed by the first predictive weight calculating unit 122 and the first predictive frequency selecting unit 124. Since the functions are the same, detailed descriptions of the N-th predictive weight calculator 126 and the N-th predictive frequency selector 128 are omitted here.

The first prediction weight calculator 122 in the candidate weight frequency frequency calculator 120 determines, for each block set, a weighted block for blocks of color planes other than the predicted reference color plane from the block set, and the color plane from the weighted block. A prediction weight for each frequency domain is calculated by stars (that is, by color planes other than the prediction reference color plane). The calculated prediction weights for each frequency domain may be transmitted to the first prediction frequency selector 124 and the optimal weighted filter determiner 130. In addition, since the image decoding apparatus according to an embodiment of the present invention and the color space prediction apparatus according to the second embodiment of the present invention receive encoded data and perform decoding, a weighted block is determined for a decoded neighboring block. Can be.

The first prediction frequency selector 124 calculates the prediction gain for each frequency domain by using the prediction weight for each frequency domain calculated by the first prediction weight calculator 122 and the frequency transform coefficients on the weighted block, and calculates the estimated gain for each frequency domain. The prediction gain is used to select a frequency domain to be used for color space prediction.

The optimal weighted filter determiner 130 receives the prediction weights for each N frequency domains and the N candidate frequency domains, calculates and compares the prediction efficiency for each, and compares the prediction weights for each frequency domain and the candidate frequency domains with the optimal prediction efficiency. Determine. Here, the prediction weights and the candidate frequency domains for each frequency domain having the optimal prediction efficiency are generated when the current block is predicted and encoded for each of the frequency weighted prediction weights and the candidate frequency domains transmitted from the candidate weighted frequency calculator. In consideration of the bit amount and the distortion value, the prediction weight and the candidate frequency domain for each frequency domain when the rate-distortion cost is the smallest are set.

The frequency domain predicting unit 140 receives the transformed residual block signal from the transforming unit 150 and receives the prediction weighting frequency and the candidate frequency domain for each frequency region having the optimal prediction efficiency selected by the optimal weighting filter determination unit 130. Color space prediction of the transformed residual block is performed.

Meanwhile, the term color space prediction may mean a series of processes for generation of a color space prediction block or generation of a frequency domain prediction residual signal.

The residual block input to the transformer 150 may be generated through inter / intra prediction of an existing image encoder. The residual block may be generated through motion prediction image generation and residual block generation, or may be generated through intra prediction image generation and residual block. Can be generated through generation. The converter 150 may calculate a transform coefficient of the residual signal block using a transform method such as a discrete cosine transform (DCT) on the input inter / intra residual block.

The first prediction weight calculator 122 may calculate a prediction weight for each frequency domain based on a correlation between frequency-space transform coefficients of the residual signal blocks of two or more color images. Since the residual signal between the color image (R, G, B signal) residual signal blocks may have a different correlation (correlation) for each frequency, an independent weight for each frequency may be set to increase the prediction efficiency between the residual signal color and the plane. Can be. The prediction weight may be calculated using a mean square error method, which is an example of implementing the present invention, and the spirit of the present invention is not limited to a specific weight calculation method.

The first prediction weight calculator 122 calculates a prediction weight from the frequency coefficients of the decoded neighboring block (ie, the first block set) based on the transform coefficients of the residual signal blocks of the decoded neighboring block, and the neighboring block. A prediction weight for each frequency domain is calculated based on the transform coefficients of the residual signal block of. Detailed operations of the first predictive weight calculation unit 122 will be described later. In the present specification, the term "decoded neighboring block" may mean a neighboring block decoded again for use as a reference block after encoding is completed when used in an encoding apparatus. It may mean a block. In addition, the term "encoded neighboring block" may be used interchangeably with the term "encoded neighboring block" in the encoding apparatus.

The first prediction frequency selector 124 may determine a frequency band to be used for prediction by using the prediction weight calculated by the first prediction weight calculator 122. That is, the first prediction frequency selector 124 selects a frequency band having a high correlation and expected to obtain an encoding gain due to frequency domain prediction.

The first prediction frequency selector 124 calculates the prediction gain for each frequency domain by applying the prediction weight for each frequency domain calculated by the first prediction weight calculator 122 to the prediction of the residual signal block, and calculates the calculated result. Based on the prediction, the frequency domain that is expected to be gained is selected to determine the frequency domain to be used for the prediction.

The frequency domain predicting unit 140 subtracts a value obtained by applying a prediction weight of a selected frequency domain to frequency coefficients of color planes other than the reference color plane from frequency coefficients of the selected frequency domain of the reference color plane of the transformed residual block. Can be performed.

The frequency domain predictor 140 performs frequency-domain color-space prediction by using the prediction weight for each frequency domain and the candidate frequency domain having the optimal prediction efficiency determined by the optimal weighted filter determiner 130. The frequency domain predictor 140 may appropriately quantize the prediction weights generated by the candidate weight frequency frequency calculator 120 to limit the calculation accuracy and the computation load of the decoder / decoder. May be omitted. The frequency domain predictor 140 removes redundancy by performing color-plane frequency domain prediction using the quantized prediction weights in the frequency domain determined by the optimal weighted filter determiner 130.

The data from which redundancy is removed by the frequency domain prediction unit 140 is compressed by an entropy encoder such as Variable Length Coding (VLC) or Context-based Adaptive Binary Arithmetic Coding (CABAC) and added to the bit string to complete the video encoding process. Decoding may be performed by performing the encoding process in reverse.

This embodiment assumes an RGB image as an input signal and is described based on one macroblock, but the format and size of the input image are not limited to this embodiment. That is, the prediction weight for each frequency domain may be calculated in any one of units of an image such as a sequence unit, a frame unit, a macroblock unit, and a subblock unit of an image to perform color space prediction.

5 is a diagram illustrating color space prediction performed by the color space prediction apparatus 100 according to the first embodiment of the present invention with respect to an RGB input image.

As shown in FIG. 5, the residual signal generator (not shown) generates a prediction image by performing intra or motion prediction on each of RGB color images, and then generates a residual image corresponding to a difference between the original image and the prediction image. To calculate. On the other hand, the conversion unit 150 performs the frequency-space conversion on the residual signal image of each of the R, G, B. It is recommended to use a DCT (Discrete Cosine Transform) or an H.264 integer transform, but the present invention is not limited thereto. Different frequency-space transforms may be used for R, G, and B residual signals. The 4x4, 8x8, or 16x16 transforms can be used for various sizes of transforms.The 4x4 transform allows 16x16 macroblocks to be transformed into 16 4x4 transform coefficient blocks. Can be transformed into 8x8 transform coefficient blocks.

The first prediction weight calculator 122 calculates a prediction weight for each optimal frequency of the RGB residual signal. The prediction weight for each frequency is determined by the relationship between the transform coefficient between the color plane used as the prediction reference color plane and the predicted color plane. For example, an image of the G plane may be used as the prediction reference color plane. The residual signal transform coefficients of the R plane and the B plane can be predicted using the residual signal transform coefficients of. Here, the prediction reference color plane is not limited to the G plane, and an R plane or a B plane may be used. On the other hand, the prediction reference color plane may be predetermined, and may also receive information on the prediction reference color plane from the reference color plane information generator (not shown). In this case, the first prediction weight calculator 122 may calculate the prediction weight by receiving the information about the prediction reference color plane, and the information about the prediction reference color plane is the first prediction frequency selector 122 and the optimal weight. It may also be transmitted to the filter determiner 130 and the frequency domain predictor 140. After encoding, the information about the prediction reference color plane may be included in the encoded data and transmitted to the decoder.

In FIG. 6, one 16x16 macroblock is composed of 16 4x4 macroblocks and performs prediction between transform coefficients in blocks of the same position of the prediction reference color plane and the other color plane. The prediction operation performed by the frequency domain predictor 140 may be expressed by Equation 1 below.

Equation 1

In Equation 1, rB (i, j) predicts the residual signal B (i, j) of the B plane with the residual signal G (i, j) of the G plane in the (i, j) th frequency domain. One frequency-domain prediction residual signal, rR (i, j) is a frequency-domain prediction residual that predicts the residual signal R (i, j) of the R plane as the residual signal of the G color plane at the (i, j) th frequency. Represent each signal. Also, F _{B / G} (G (i, j)) is a prediction function for predicting the B plane using the G plane and F _{R / G} (G (i, j)) uses the G plane to determine the R plane. As a prediction function for predicting, each may be represented by a linear function as in Equation 2.

Equation 2

In equation (2). w _{B / G} (i, j) and w _{R / G} (i, j) are generated at frequencies at the (i, j) position of the B-plane and R-plane transform coefficient blocks, generated using decoded neighboring blocks. Each of the prediction weights may be represented and calculated using a mean square error method as shown in Equation 3 below.

Equation 3

In Equation 3, the E {} function is a function representing an expected value, and the prediction weights calculated by the first prediction weight calculator 122 may be adaptively calculated for all blocks, and different weights for each frequency position may be applied. Can be calculated. In addition, the prediction weight is preferably determined adaptively for each macroblock, but is not limited thereto. In addition, the prediction function is not limited to the linear function as in Equation 2.

In the video encoding apparatus, as shown in Equation 3, the prediction weight may be adaptively calculated for each frequency domain for each macroblock. However, since the DCT coefficient of the current macroblock is not known in advance, the decoder may perform prediction for each macroblock. And information such as weights need to be transmitted. Since many bits are required to encode this information, the present invention can calculate the frequency domain and the weight to be used for prediction by the current macroblock using the information of the neighboring blocks already encoded.

Accordingly, the first prediction weight calculator 122 may have a function of determining a neighboring block for predicting the weight of the current block, and a function of determining the prediction weight of the current block using the determined blocks.

As shown in FIG. 2 and FIG. 3 together showing the macroblock to be encoded and the already encoded macroblock row, a neighboring block that is already encoded to calculate a prediction weight of the current macroblock located in the k-th macroblock row or The DCT coefficients of the quantized G / B / R residual signal planes of the macroblocks of the n macroblock rows may be calculated based on the above-described prediction weight determination method. This is the result of considering the similarity between the current macroblock and the decoded neighboring macroblock and the minimum data acquisition for statistical stability.

In this case, since DCT coefficient distributions may be different for each mode such as intra_4x4, intra_8x8, and intra_16x16, the frequency domain and the prediction weight to be used for prediction may be calculated separately for each macroblock mode. That is, macroblocks are classified into modes such as intra_4x4 mode, intra_8x8 mode, and intra_16x16 mode in the previous macroblock row, and then, for example, macroblocks encoded in intra_4x4 mode are displayed as intra4x4 mode. A prediction weight may be calculated by collecting only encoded neighboring blocks (or weighted blocks). In the intra_8x8 mode and the intra_16x16 mode, the prediction weights may be calculated from the neighboring blocks of the same mode. Therefore, the block referenced by the prediction weight calculator 120 may be a block encoded in the same mode as the prediction mode of the current block.

The first prediction weight calculator 122 may calculate a prediction weight using Equation 3 by using the DCT coefficients of the determined blocks.

After calculating the prediction weights, the first predictive frequency selector 124 selects a frequency domain in which a gain can be obtained due to the prediction. The first prediction frequency selector 124 calculates a gain for each frequency, when the prediction is performed using the prediction weights calculated for each frequency domain. This may be calculated by Equation 4.

Equation 4

In Equation 4, Gain _{B / G} (i, j) and Gain _{R / G} (i, j) are predicted gains of (i, j) frequency positions in the B plane and (i, j) frequency positions in the R plane, respectively. It means the predicted gain of. In addition, t means all blocks in the first block set selected to calculate the prediction weight, that is, if there are three blocks in the first block set, the prediction gains are calculated and summed for all three blocks. In addition, T ₁ , T ₂ means a threshold value having an arbitrary value for prediction frequency selection.

After calculating the gain as described above, the first prediction frequency selector 124 compares the case where the prediction is performed by the frequency domain with the case where the prediction is not performed, and selects only the frequency domain having the gain when the prediction is performed. . This may be represented by Equation 5.

Equation 5

In Equation 5, B _t (i, j) and R _t (i, j) denote frequency coefficients of a block in the t th first block set in the B plane and the R plane at the position (i, j), respectively. do. The prediction gain is obtained for each block t in the first block set for each (i, j) frequency domain, and then compared with the sum of the absolute values of the frequency coefficients at the positions (i, j) of the blocks in the first block set. do. As a result of the comparison, only the frequency domain in the case of gain (that is, the case where the gain of gain (Gain _{B / G} (i, j), Gain _{R / G} (i, j)) is smaller) is used. do. That is, only the (i, j) frequency component in the case of satisfying Equation 5 performs color space prediction. In this embodiment, Gain _{B / G} (i, j) and Gain _{R / G} (i, j) are expressed as predictive gains, but the actual Gain _{B / G} (i, j) and Gain _{R / G} (i, j The smaller the value of), the greater the likelihood of gain when color space prediction is performed, and thus, the greater the likelihood that the corresponding (i, j) frequency is selected as the prediction frequency.

The frequency domain predictor 140 performs color space prediction by using the prediction weight and the predicted frequency domain selected by the optimal weighted filter determiner 130.

The frequency domain predictor 140 may include a weight quantizer (not shown) and a predictor (not shown). The weight quantization unit (not shown) is a part for reducing the computational complexity of the prediction weight, and is used to perform prediction by integer operation to prevent the mismatch of prediction weight between the encoder and the decoder. Case may be omitted.

Next, an embodiment of using a weight quantization unit (not shown) will be described. Since most prediction weights can exist as floating point numbers between -1 and 1, the weighting factor is integerized using a scaling factor and then divided by an integer in order to perform an integerization operation. In addition, since the division operation requires a large amount of operations even for integer operations, it is necessary to scale it to 2 ^m (m is a positive integer) to use a shift operation. This may be represented by Equation 6.

Equation 6

In Equation 6, a is a prediction weight having an actual floating point, and floor (b) means an integer not exceeding b.

The final prediction may be made as shown in Equation 7 by using the above prediction weights. (Here, the prediction frequencies determined in the R and B color planes may be independent of each other.)

Equation 7

As shown in Equation 7, the frequency domain predictor 140 removes the redundancy between the residual signal of the reference color plane and the transform coefficients of the prediction residual signal for each color space other than the reference color plane among the residual blocks transformed by the converter 150. Entropy-encode the encoded data to complete the video code bit sequence. If color space prediction is used at the (i, j) th frequency, frequency domain prediction residuals remain for the B and R color planes (that is, color spaces other than the reference color plane). If signals rB (i, j) and rR (i, j) are entropy coded and color space prediction is not used at the (i, j) th frequency, instead of residual signals rB (i, j) and rR (i, j) Residual signals B (i, j) and R (i, j) may be encoded.

7 is a diagram illustrating before (a) and after (b) resetting the filter type.

As shown in FIG. 7, when the set of macroblocks according to the weight filter types of the macroblocks belonging to the upper macroblock row of the current block is part of the N weighted candidate block sets, the color space prediction apparatus 100 according to an embodiment of the present invention. ) May include a row prediction detector 160 and a filter type adjuster 170.

When the color space prediction of one macroblock row is completed, the filter type may be reset for the macroblock row. If the weighted candidate block set can be composed of two sets as shown in FIG. 4 except for the set of neighboring blocks as shown in FIG. 3, each macroblock uses weights (filters) generated from each weighted candidate block set in color space prediction. Color space prediction can be performed. Thus, color space prediction may be performed using one of two prediction filters.

As shown in FIG. 7, the (k-1) th macroblock row is divided into two block sets MB_A and MB_B according to the filter type of the block set in the (k-2) th macroblock row referred to for performing color space prediction. In the case of division, it is assumed that weights that can be generated by each block set MB_A and MB_B are filters A and B, respectively.

As shown in (a) of FIG. 7, the (k-1) -th block may be divided into a macroblock predicted by the filter A and a macroblock predicted by the filter B and decoded. When the color space prediction of the entire (k-1) -th macroblock row is initially completed, the multi-prediction filter calculation unit collects the blocks predicted by the filter A and the blocks predicted by the filter B, respectively, to generate new multi-prediction filters. Calculate. These are referred to as filter A1 and filter B1, respectively, and can be calculated by Equation 3.

After the color space prediction of the (k-1) -th macroblock row is completed, the filter type adjusting unit 170 performs frequency based on the conversion coefficient of the macroblock for each set of macroblocks for each filter type of the (k-1) -th macroblock row. Type-specific weights (filters A1 and B1), which are predictive weights for each region, are calculated by Equation 3, and the type-specific weights have the optimal filter efficiency for all macroblocks in the (k-1) th macroblock row. Resets the filter type for each macroblock.

Here, the optimal filter rate is selected in the case where the rate-distortion cost is the smallest in consideration of the bit amount and distortion generated when the current block is predicted and encoded for each type of weight (filter A1 and filter B1). As such, the filter type is reset for every macroblock in the (k-1) th macroblock row. At this time, the type of each macroblock is determined according to the selected filter type. If the filter type is changed, the filter type is reset.

That is, when filter A1 is selected, the filter type of the macroblock is set to filter A.

FIG. 7B shows the filter type of the macroblock changed after the filter type readjustment is performed.

It can be seen that the filter arrangement of FIG. 7B and the filter arrangement of FIG. 7A are different.

When the filter type adjusting unit 170 resets the filter type and determines that the number of reset (ie, changed) filter types is smaller than the preset number (threshold), the filter type adjusting unit 170 determines that the final filter type is higher than the preset number (threshold). Repeat the reset of the filter type.

As shown in FIG. 8, the color space prediction apparatus 800 according to the second embodiment of the present invention includes a weight candidate candidate selecting unit 810, a candidate weight frequency calculating unit 820, an optimal weighting filter determining unit 830, and the like. Frequency domain recovery unit 840 is included. In some cases, a row prediction detector 860 and a filter type adjuster 870 may be added.

The weight candidate selection unit 810 determines a set of N weight candidate blocks from the transform coefficient blocks of the decoded residual signal.

The candidate weighted frequency calculating unit 820 calculates the prediction weight for each frequency domain based on the transform coefficients of the weighted candidate blocks for each of the N weighted candidate blocks, calculates the prediction gain for each frequency domain from the prediction weight for the frequency domain, and calculates the frequency domain. A candidate frequency region to be used for color space prediction is selected by using the predictive gain of each star.

The optimal weighted filter determiner 830 receives the prediction weights for each N frequency domains and the N candidate frequency domains, calculates and compares the prediction efficiency for each, and compares the prediction weights and the candidate frequency domains for each frequency domain having an optimal prediction efficiency. Determine.

The frequency domain reconstruction unit 840 reconstructs the residual block transformed from the color space prediction block by using the prediction weight for each frequency domain and the candidate frequency domain having the optimal prediction efficiency.

The row prediction detector 860 checks whether the color space prediction of one macroblock row is completed.

After the color space prediction of one macroblock row is completed, the filter type adjusting unit 870 performs the type weighting value, which is a prediction weight for each frequency domain, based on the transform coefficient of the macroblock for each set of macroblocks for each filter type of one macroblock row. Calculate and reset the filter type for the macroblock according to the type-specific prediction weights with the optimal filter efficiency for all macroblocks in one macroblock row.

In the color space prediction apparatus 800 according to the second embodiment of the present invention, the weight candidate candidate selector 910, candidate weight frequency calculator 820, optimal weighted filter determiner 830, and row prediction detector 860 ) And the filter type adjusting unit 870 may include a weighted candidate selection unit 110, a candidate weighted frequency calculating unit 120, an optimal weighted filter determining unit 130, and the color space prediction apparatus 100 according to the first embodiment. Since the row prediction detection unit 160 and the filter type adjustment unit 170 are the same, detailed description thereof will be omitted.

Here, the prediction weights for each frequency domain and the candidate frequency domain having the optimal prediction efficiency calculated by the optimal weighting filter determiner 830 may be transmitted to the frequency domain reconstruction unit 840.

The frequency domain reconstruction unit 840 reconstructs the residual block transformed from the color space prediction block by using the prediction domain weight and the candidate frequency domain for each frequency region having the selected optimal prediction efficiency.

The frequency domain reconstructor 840 converts the frequency coefficients of the selected frequency domain of the reference color plane of the color space prediction block and the values obtained by applying the prediction weights of the selected frequency domain to the frequency coefficients of the color planes other than the reference color plane. The residual block can be restored.

Here, when restoring the transformed residual block by transforming Equation 1, when restoring the transformed residual block of the B plane, [B (i, j) = rB (i, j) + F _{B / G} (G (i , j))] can be used, and the R plane can be obtained similarly to the B plane.

That is, the operation of the frequency domain reconstructor 840 may be performed in the reverse order of the operation of the frequency domain predictor 140, and may be applied to the first block set, the prediction weight for each frequency domain, the prediction gain for each frequency domain, and the color space prediction. Since the selection of the frequency domain to be used is described in the color space prediction apparatus 100 according to the first embodiment of the present invention, a detailed description thereof will be omitted.

In the color space prediction apparatus 800 according to the second embodiment of the present invention, the first predictive weight calculation unit 822 targets blocks of color planes other than the prediction reference color plane for the decoded neighboring blocks of the current block. While the weighted block is determined with respect to the first block, the first prediction weight calculator 122 in the color space prediction apparatus 100 according to the first embodiment of the present invention targets the decoded neighboring block of the current block. The operation of the first predictive weight calculation unit 822 according to the second embodiment of the present invention differs only in that the weighted block is determined for the blocks of the color plane other than the color plane. It is the same as the predictive weight calculation unit 122.

The frequency domain reconstructor 840 uses the prediction weights F _{B / G} (G (i, j)) and F _{R / G} (G (i, j)) of the frequency domain selected in Equation (1). It is possible to reconstruct the transformed residual block by calculating B (i, j) and R (i, j) from the prediction blocks rB (i, j) and rR (i, j).

The image encoding apparatus 900 according to an embodiment of the present invention includes a predictor 910, a subtractor 920, a transformer 930, a color space predictor 932, and a color space decompressor 934. It may be configured to include a scanner (Scanner, 940), an encoder (Encoder, 950), an inverse transformer (960), an adder (970), and a filter (980).

The input image to be encoded may be input in block units, and the block may be a macroblock. In one embodiment of the present invention, the shape of the macroblock is P × Q, where P is 2 ^m (m is an integer of 1 or more), and Q is 2 ⁿ (where n is an integer of 1 or more). It may be in various forms such as. In addition, different types of blocks may be used for each frame to be encoded, and when a macroblock has various forms such as P × Q, information about a block type, which is information about this, is encoded for each frame and encoded by the image decoding apparatus. When decoding the decoded data, the shape of the block of the frame to be decoded may be determined. Determination of which type of block to use can be made by selecting the type of block that has the best efficiency by encoding the current frame into various types of blocks, or by selecting the type of block according to the analyzed characteristics by analyzing the characteristics of the frame. have. To this end, the image encoding apparatus 900 may further include a block type determiner (not shown) which determines the block type and encodes information about the block type in the encoded data.

The predictor 910 predicts the current block for each color plane to generate a predicted block. That is, the predictor 910 predicts a pixel value of each pixel of the current block to be encoded in the image and generates a predicted block having a predicted pixel value of each pixel predicted. do. Here, the predictor 910 may predict the current block by using intra prediction or inter prediction.

The subtractor 920 subtracts the prediction block from the current block to generate a residual block. That is, the subtractor 920 calculates a difference between the pixel value of each pixel of the current block to be encoded and the predicted pixel value of each pixel of the prediction block predicted by the predictor 910 to obtain a residual signal in the form of a block. Creates a residual block with

When the transformer 930 transforms the residual block, the transformation process may be included in the quantization process. In this case, the transformation is completed when the quantization is completed. Here, the transform method includes a spatial signal such as a Hadamard transform and a discrete cosine transform based integer transform (hereinafter, referred to as an integer transform) to the frequency domain. Transformation techniques may be used, and various quantization techniques such as Dead Zone Uniform Threshold Quantization (DZUTQ) or Quantization Weighted Matrix (DZUTQ) are used as quantization schemes. Can be.

The color space predictor 932 performs color space prediction based on the transform coefficients of the blocks in the weighted candidate block set for each of the N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. Select the prediction weights and candidate frequency domains for each frequency domain to be used, and select the prediction weights for each frequency domain and the candidate frequency domains having the optimal prediction efficiency among the candidate frequency domains and the prediction weights for each of the N weight candidate blocksets. To generate a color space prediction block for the transformed residual block. Here, the color space predictor 932 may predict the color space using the color space prediction apparatus 100 according to the first embodiment of the present invention. Since the detailed description thereof has been described above in the description of the color space prediction apparatus 100 according to the first embodiment of the present invention, the detailed description thereof will be omitted.

Meanwhile, the transformer 930 may perform quantization at the same time, but may also quantize the color space prediction block by the color space predictor 932 instead of the transformer 930.

The coefficients of the color space prediction block generated by the color space predictor 932 are scanned to generate a coefficient sequence. In this case, the scanning method considers characteristics of a transform technique, a quantization technique, and a block (macroblock or subblock), and the scanning order may be determined such that the scanned coefficient sequence has a minimum length. In FIG. 9, the scanner 940 is illustrated and described as being implemented independently of the encoder 950, but the scanner 940 may be omitted, and its function may be integrated into the encoder 950.

The encoder 950 encodes the coefficients of the color space prediction block generated by the color space predictor 932.

The encoder 950 encodes a coefficient string generated by scanning coefficients of the color space prediction block generated by the color space predictor 932 to generate encoded data, or encodes a coefficient string generated by scanning by the scanner 940. To generate encoded data.

As the encoding technique, an entropy encoding technique may be used, but various encoding techniques may be used without being limited thereto. In addition, the encoder 950 may include not only a bit string encoding the coefficient string of the color space prediction block, but also various pieces of information necessary to decode the encoded bit string in the encoded data. Here, the various pieces of information necessary to decode the encoded bit strings include information about a block type, information about an intra prediction mode when the prediction mode is an intra prediction mode, and information about a motion vector when the prediction mode is an inter prediction mode. Information, information on transform and quantization types, and the like, but various other information.

The color space decompressor 934 predicts the color space based on the transform coefficients of the blocks in the weighted candidate block set for each of the N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. Select the prediction weights and the candidate frequency domains for each frequency domain to be used for the frequency domain, and the prediction weights for each frequency domain and the candidate frequency domains having the optimal prediction efficiency among the candidate frequency domains Reconstruct the residual block transformed from the color space prediction block using.

Here, the color space reconstructor 934 may reconstruct the transformed residual block using the color space prediction apparatus 800 according to the second embodiment of the present invention. Since the detailed description thereof has been described above in the description of the color space prediction apparatus 800 according to the second embodiment of the present invention, the detailed description thereof will be omitted.

The inverse transformer 960 reconstructs the residual block by inverse transforming the transformed residual block restored by the color space reconstructor 934. If the quantization is also performed in the transformer 930, the inverse transformer 960 may perform inverse transformation after inverse quantization and may be performed by inversely performing a transform process and a quantization process performed by the transformer 930. If quantization is also performed in the converter 1030 to perform inverse transform after inverse quantization in the inverse transformer 960, the transform may be described as being replaced with transform and quantization, and the inverse transform is replaced with inverse quantization and inverse transform. Can be explained.

The inverse transformer 960 receives information about the transformation from the transformer 930 and receives the information about the transformation to perform the inverse transformation of the transformer 930 and perform inverse transformation ( Or inverse quantization and inverse transformation).

The adder 970 reconstructs the current block by adding the prediction block predicted by the predictor 910 and the residual block generated by the inverse transformer 960.

The filter 980 filters the current block reconstructed by the adder 970. In this case, the filter 980 reduces blocking effects occurring at the block boundary or the transform boundary by transform and quantization of blocks in an image.

The image decoding apparatus 1000 according to an exemplary embodiment of the present invention includes a decoder 1010, an inverse scanner 1020, an inverse transformer 1030, an adder 1040, a predictor 1050, and a filter 1060. Can be configured. In this case, the inverse scanner 1020 and the filter 1060 are not necessarily included and may be selectively omitted depending on the implementation manner. When the inverse scanner 1020 is omitted, the function is integrated in the decoder 1010. Can be implemented.

The decoder 1010 decodes the encoded data to decode the color space prediction block. That is, the decoder 1010 may reconstruct the color space prediction block by decoding the encoded data. When the function of the scanner 940 is integrated with the encoder 950 in the image encoding apparatus 900, the image decoding apparatus Since the inverse scanner 1020 is also omitted at 1000 and its function is integrated and implemented in the decoder 1010, the decoder 1010 may rescan the encoded data to reconstruct the color space prediction block.

In addition, the decoder 1010 may decode the encoded data to decode or extract not only the color space prediction block but also information necessary for decoding. The information necessary for decoding refers to information necessary for decoding the coded bit string in the encoded data. For example, information about a block type, information about an intra prediction mode when the prediction mode is an intra prediction mode, and an inter prediction mode In the case of the prediction mode, the information may be information on a motion vector, information on a transform and quantization type, or the like.

When information about a block type is transmitted and input, information about a block type may be transmitted to the inverse transformer 1030 and the predictor 1050, and information about a transform type (or a transform and quantization type) may be transferred to the inverse transformer 1030. In addition, information necessary for prediction, such as information about an intra prediction mode and information about a motion vector, may be transmitted to the predictor 1050. Information about the quantization type may also be transmitted to the color space reconstructor 1070.

When the inverse scanner 1020 restores and transmits the color space transform coefficient sequence from the decoder 1010, the inverse scanner 1020 restores the color space prediction block by inversely scanning the color space transform coefficient sequence.

The inverse scanner 1020 inversely scans the extracted coefficient sequence by various inverse scanning methods such as inverse zigzag scan to generate a color space prediction block. In this case, the inverse scanning method may obtain information about the size of the transform from the decoder 1010 and generate a residual block using the corresponding inverse scanning method. Here, the inverse scanning method of the inverse scanner 1020 according to the transform and quantization type is the same as or similar to that of the scanner 940 inversely scanning the coefficients of the color space prediction block. Detailed description thereof will be omitted.

The color space reconstructor 1070 predicts the color space based on the transform coefficients of the blocks in the weighted candidate block set for each of the N (N is natural numbers of 2 or more) weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. Select the prediction weights and the candidate frequency domains for each frequency domain to be used for the frequency domain, and the prediction weights for each frequency domain and the candidate frequency domains having the optimal prediction efficiency among the candidate frequency domains Reconstruct the residual block transformed from the color space prediction block using.

The color space reconstructor 1070 may reconstruct the residual block converted from the color space prediction block by using the color space prediction apparatus 800 according to the second embodiment of the present invention. Detailed description thereof has been described above in the description of the color space prediction apparatus 800 according to the second embodiment of the present invention, and since the same operation as that of the color space decompressor 1034 may be performed, a detailed description thereof will be omitted.

The inverse transformer 1030 inversely transforms the transformed residual block to be recovered to restore the residual block. In this case, the inverse transformer 1030 may inversely transform the residual block transformed according to the transformation type. Here, the method of inversely transforming the transformed residual block according to the transform type by the inverse transformer 1030 is the same as or similar to inversely performing the process of transforming the transformed block according to the transform type in the converter 930 of the image encoding apparatus 900. Therefore, detailed description of the inverse transform method is omitted.

The predictor 1050 predicts the current block to generate a predicted block.

The predictor 1050 determines the size and shape of the current block according to the block type identified by the information on the block type, and predicts the current block by using an intra prediction mode or a motion vector identified by the information required for prediction. A prediction block can be generated. In this case, the predictor 1050 is the same or similar to the predictor 910 of the image encoding apparatus 900. The predictor 1050 divides the current block into subblocks and combines predicted subblocks generated by predicting the divided subblocks to predict the predicted block. Can be generated.

The adder 1040 adds the residual block reconstructed by the inverse transformer 1030 and the predictive block generated by the predictor 1050 to reconstruct the current block.

The filter 1060 filters the current block reconstructed by the adder 1040, and the reconstructed filtered current block is accumulated in units of pictures and stored in a memory (not shown) as a reference picture to be stored in the next block or in the predictor 1050. It can be used when predicting the next picture.

Since the method of performing the filtering by the filter 1060 is the same as or similar to that of the deblocking filtering by the filter 980 of the image encoding apparatus 900, a detailed description of the filtering method is omitted.

Meanwhile, the image encoding / decoding apparatus according to an embodiment of the present invention may be implemented by connecting the encoded data output terminal of the image encoding apparatus 900 of FIG. 9 to the encoded data input terminal of the image decoding apparatus 1000 of FIG. 10.

In an apparatus for encoding / decoding an image, the apparatus for encoding / decoding an image according to an embodiment of the present invention generates a prediction block by predicting a current block for each color plane and subtracts the residual block by subtracting the prediction block from the current block. Generating, transforming the residual block, and for each N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal, based on the transform coefficients of the blocks in the weighted candidate block set. Prediction weights and candidate frequencies for each frequency domain are selected by selecting the prediction weights and candidate frequency domains for each frequency domain to be used for prediction and having optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weighted candidate blocksets. Create a color space prediction block for the transformed residual block using the domain. An image encoder (which can be implemented using the image encoding apparatus 900) and the encoded data to decode the color space prediction block, and decode the color space prediction block and select a transform coefficient block of the decoded residual signal. For each of N (N is a natural number of 2 or more) weighted candidate block sets, a prediction weighted candidate frequency band and candidate frequency domain to be used for color space prediction are selected based on the transform coefficients of the blocks in the weighted candidate block set, and N weighted candidate blocks Reconstruct the residual blocks transformed from the color space prediction blocks by using the frequency domain prediction weights and the candidate frequency domains having the optimal prediction efficiency among the frequency domain prediction weights and the candidate frequency domains calculated for each set, and convert the residuals. Invert the block to restore the remaining blocks, and predict the current block. Generating a prediction block, by adding the prediction block and the reconstructed residual block comprises a video decoder (which may be implemented by using the image decoding apparatus 1000) to reconstruct the current block.

1 to 11, the color space prediction method according to the first embodiment of the present invention includes determining N weighted candidate blocksets from transform coefficient blocks of a decoded residual signal (S1102), N Calculating prediction weights for each frequency domain based on the transform coefficients of the weighted candidate blocks for each of the weighted candidate blocks (S1104), calculating prediction gains for each frequency domain from the prediction weights for each frequency domain (S1106), and frequency domain Selecting a candidate frequency region to be used for color space prediction using the prediction gain for each step (S1108), receiving the prediction weights for each N frequency domains and the N candidate frequency domains, calculating and comparing the prediction efficiency for each, Determining the prediction weight and the candidate frequency domain for each frequency domain having a prediction efficiency (S1110), the frequency zero having an optimal prediction efficiency A frequency domain prediction step S1112 of generating a color space prediction block for the residual block transformed using the inverse prediction weights and the candidate frequency domain may be included.

Here, in step S1102, the operation of the weight candidate selection unit 110 is performed, in steps S1104, S1106, and S1108, the operation of the candidate weight frequency frequency calculating unit 120 is performed. Since the frequency domain prediction step S1112 may correspond to the operation of the frequency domain predictor S140 in operation of the optimal weight filter determiner 130, a detailed description thereof will be omitted.

In the color space prediction method according to the second embodiment of the present invention, determining a set of N weighted candidate blocks from the transform coefficient blocks of the decoded residual signal (S1202), and weighting candidate blocks for each of the N weighted candidate blocks. Computing the prediction weight for each frequency domain based on the conversion coefficient of (S1204), Computing the prediction gain for each frequency domain from the prediction weight for each frequency domain (S1206), Predicting the color space using the prediction gain for each frequency domain In step S1208, the candidate frequency domain to be used is received, the prediction weights for each of the N frequency domains and the N candidate frequency domains are calculated, and the prediction efficiency is calculated and compared for each of the frequency domains. Determining the candidate frequency domain (S1210), and prediction weights and candidate frequencies for each frequency domain having an optimal prediction efficiency Using an area may include a frequency-domain restoration step of restoring the residual block from the converted color space prediction block (S1212).

Here, in step S1202, the operation of the weight candidate selection unit 810 is performed, in steps S1204, S1206, and S1208, the operation of the candidate weight frequency frequency calculating unit 820 is performed. Since the frequency domain restoring step S1212 may correspond to the operation of the frequency domain restoring unit S840, the detailed description thereof will be omitted.

1 to 13, the image encoding method according to an embodiment of the present invention includes generating a prediction block by predicting a current block for each color plane (S1302), and subtracting the prediction block from the current block. A transform step (S1304) of generating a residual block and transforming the residual block, for each of the N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal, based on the transform coefficients of the blocks in the weighted candidate block set. Selecting the prediction weights and candidate frequency domains for each frequency domain to be used for prediction (S1306), for each frequency domain having an optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weighted candidate blocksets Computing a prediction weight and a candidate frequency domain (S1308), for each frequency domain having an optimal prediction efficiency And a side weight, and step (S1312) for coding a color space that step (S1310), and color space prediction block to generate a color space prediction block candidates for the frequency domain transform to the residual block with.

Here, step S1302 corresponds to the operation of the predictor 910, step S1304 refers to the operations of the subtractor 920 and the converter 930, step S1306, step S1308, and color space prediction step S1310. ) May correspond to the operation of the color space predictor 932 and operation S1312 may correspond to the operation of the encoder 950, and thus a detailed description thereof will be omitted.

1 to 14, the image decoding method according to an embodiment of the present invention includes decoding the color space prediction block by decoding the encoded data (S1402), and transform coefficient block of the decoded residual signal. Selecting the prediction weighted region and the candidate frequency region for each frequency domain to be used for the color space prediction based on the transform coefficients of the blocks in the weighted candidate block set for each of the N weighted candidate block sets selected from S1404, and the N weighted candidate blocks Calculating the prediction weights for each frequency domain and the candidate frequency domains having the best prediction efficiency among the frequency domain prediction weights and the candidate frequency domains calculated for each set (S1406), and the prediction weights for each frequency domain with the optimal prediction efficiency. And reconstructing the residual block transformed from the color space prediction block using the candidate frequency domain. Color space reconstruction step (S1408), inverse transform step of reconstructing the transformed residual block (S1410), generating a prediction block by predicting the current block (S1412) and reconstructing the residual block and the prediction block And adding to restore the current block (S1414).

Here, step S1402 is an operation of the decoder 1010, step S1404, step S1406, and step S1408 are operations of the color space decompressor 1070, and step S1410 is an inverse transformer 1030. In operation S1412, operation S1412 may correspond to operation of the predictor 1050, operation S1414 may correspond to operation of the adder 1040, and thus a detailed description thereof will be omitted.

An image encoding / decoding method according to an embodiment of the present invention may be realized by combining a video encoding method according to an embodiment of the present invention of FIG. 13 and an image decoding method according to an embodiment of the present invention of FIG. 14. Can be.

In an image encoding / decoding method according to an embodiment of the present invention, a prediction block is generated by predicting a current block for each color plane, and a residual block is generated by subtracting the prediction block from the current block, transforming the residual block, and ups and downs. Prediction weights and candidates for each frequency domain to be used for color space prediction based on the transform coefficients of the blocks in the weighted candidate block set for each of the N weighted candidate block sets selected from the transform coefficient blocks of the called residual signal. For the residual block transformed by using the frequency domain prediction weights and candidate frequency domains having the optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains selected for each of the N weighted candidate blocksets. Generating the color space prediction block and encoding the color space prediction block Decode the color space prediction block by decoding the encoded data, and apply the transform coefficients of the blocks in the weighted candidate block set to each of the N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. Selecting the prediction weights and candidate frequency domains for each frequency domain to be used for the color space prediction based on the frequency domains having the optimal prediction efficiency among the frequency weighted prediction weights and candidate frequency domains calculated for each of the N weighted candidate blocksets. Reconstruct the residual block transformed from the color space prediction block by using the prediction weights and the candidate frequency domain, restore the residual block by inversely transforming the transformed residual block, generate the prediction block by predicting the current block, and restore the residual Reconstruct the current block by adding a block and the prediction block It may be made, including the steps:

As described above, according to an embodiment of the present invention, in order to efficiently encode the motion vector of the current block, the context of the motion vector is generated based on the motion vector correlation of the neighboring block, and the candidate motion vector is generated as the context of the neighboring block. By adaptively adapting to, the encoding performance of the motion vector of the current block is greatly improved, thereby improving the encoding performance of the video compression apparatus or the quality of the reconstructed video.

In the above description, it is described that all the components constituting the embodiments of the present invention are combined or operated in one, but the present invention is not necessarily limited to these embodiments. In other words, within the scope of the present invention, all of the components may be selectively operated in combination with one or more. In addition, although all of the components may be implemented in one independent hardware, each or all of the components may be selectively combined to perform some or all functions combined in one or a plurality of hardware. It may be implemented as a computer program having a. Codes and code segments constituting the computer program may be easily inferred by those skilled in the art. Such a computer program may be stored in a computer readable storage medium and read and executed by a computer, thereby implementing embodiments of the present invention. The storage medium of the computer program may include a magnetic recording medium, an optical recording medium, a carrier wave medium, and the like.

In addition, the terms "comprise", "comprise" or "having" described above mean that the corresponding component may be inherent unless specifically stated otherwise, and thus excludes other components. It should be construed that it may further include other components instead. All terms, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. Terms used generally, such as terms defined in a dictionary, should be interpreted to coincide with the contextual meaning of the related art, and shall not be interpreted in an ideal or excessively formal sense unless explicitly defined in the present invention.

The above description is merely illustrative of the technical idea of the present invention, and those skilled in the art to which the present invention pertains may make various modifications and changes without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention but to describe the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The protection scope of the present invention should be interpreted by the following claims, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of the present invention.

As described above, according to the exemplary embodiment of the present invention, video data can be directly and effectively compressed without a conventional color conversion process. In addition, coding efficiency may be further improved by removing overlapping information between color components depending on encoding modes by using correlation between image components.

In addition, if the direct encoding is performed in the color space region of the original image, there is no loss of image quality such as color distortion that occurs when converting to another region, and thus, digital cinema and digital archives requiring high quality image information It is suitable for applications such as (Digital Archive), so it is highly industrially applicable.

CROSS-REFERENCE TO RELATED APPLICATION

If this patent application claims priority under US Patent § 119 (a) (35 USC § 119 (a)) to patent application No. 10-2010-0071238 filed with Korea on July 23, 2010, All of which is incorporated herein by reference. In addition, if this patent application claims priority for the same reason as above for a country other than the United States, all the contents thereof are incorporated into this patent application by reference.

Claims

An apparatus for encoding / decoding a video,

A prediction block is generated by predicting a current block for each color plane, and a residual block is generated by subtracting the prediction block from the current block, the residual block is transformed, and a selected N (from a transform coefficient block of a decoded residual signal). N is a weighted candidate block set for each frequency domain to be used for color space prediction based on the transform coefficients of the blocks in the weighted candidate block set for each of the weighted candidate block sets, and the N weighted candidate blocks. A color space prediction block is generated for the transformed residual block by using the prediction weighted frequency and the candidate frequency domain having the optimal prediction efficiency among the frequency domain prediction weights and the candidate frequency domains calculated for each set. An image encoder for encoding a color space prediction block; And

Decoding the coded data to decode the color space prediction block, and transform coefficients of blocks in the weighted candidate block set with respect to each of N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. Select a prediction weight region and a candidate frequency region for each frequency domain to be used for color space prediction based on the frequency domain having the optimal prediction efficiency among the predicted weight region and the candidate frequency region for each of the N weighted candidate blocksets Reconstruct a residual block transformed from the color space prediction block by using a prediction weight and a candidate frequency domain for each star, generate a prediction block by predicting a current block, and add the reconstructed residual block and the prediction block to the current block. Video decoder to restore

Image encoding / decoding apparatus comprising a.
In the apparatus for encoding a video,

A predictor generating a prediction block by predicting the current block for each color plane;

A subtractor for generating a residual block by subtracting the prediction block from the current block;

A transformer for transforming the residual block;

Prediction weights for each frequency domain to be used for color space prediction based on the transform coefficients of blocks in the weighted candidate block set for each of the weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. And selecting the candidate frequency domain and converting the prediction weighted frequency and the candidate frequency domain by the frequency domain with the optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weighted candidate blocksets. A color space predictor for generating a color space prediction block for the residual block; And

An encoder for encoding the color space prediction block

An image encoding apparatus comprising a.
The method of claim 2,

The prediction weight for each frequency domain is

And an image encoding device calculated by a correlation between the reference color plane and the remaining color planes.
The method of claim 2,

The prediction weight for each frequency domain is

The image encoding apparatus is calculated in any one of a sequence unit, a frame unit, a macroblock unit, and a subblock unit of an image.
The method of claim 2,

The color space predictor,

The color space prediction block is generated by subtracting a value obtained by applying a prediction weight of the selected frequency domain to a frequency coefficient of a color plane other than the reference color plane from a frequency coefficient of the selected frequency domain of the reference color plane of the transformed residual block. And a video encoding apparatus.
The method of claim 2,

The color space predictor,

A weighted candidate selection unit for determining a set of N-weighted candidate blocks from the transform coefficient blocks of the decoded residual signal;

For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated using a transform coefficient of a block in the weighted candidate block set, and a prediction gain for each frequency domain is calculated using the prediction weights for each frequency domain. A candidate weight frequency calculator for selecting a candidate frequency domain to be used for color space prediction by using the prediction gain for each frequency domain;

An optimum weighting filter determiner for receiving prediction weights and N candidate frequency domains for each of the N frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights and the candidate frequency domains for each frequency domain having an optimal prediction efficiency; And

A frequency domain prediction unit for generating a color space prediction block for the transformed residual block by using the prediction weights for each frequency domain and the candidate frequency domain having the optimal prediction efficiency.

An image encoding apparatus comprising a.
The method of claim 6,

The N weight candidate blocksets are

And a set of neighboring blocks and a decoded M (M is one or more natural numbers) macroblock row set.
The method of claim 6,

The N weight candidate blocksets are

And a macroblock for each filter type of the macroblock belonging to the decoded macroblock row.
The method of claim 8,

The color space predictor,

A row prediction detector for checking whether color space prediction of one macroblock row is completed; And

After the color space prediction of the one macroblock row is completed, the weight of each type, which is a prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of macroblocks of each filter type of the one macroblock row. Filter type adjustment unit for resetting the filter type for the macroblock according to the type-specific prediction weights having the optimal filter efficiency for all macroblocks in the one macroblock row

The image encoding apparatus further comprises.
The method of claim 9,

The optimal filter rate is

And a rate-distortion cost is smallest in consideration of the amount of bits and distortions generated when the current block is predicted and encoded for each type-weighted value.
The method of claim 9,

The filter type adjustment unit,

And resetting the filter type if the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.
The method of claim 6,

The prediction weight for each frequency domain and the candidate frequency domain having the optimal prediction efficiency are

Frequency when the rate-distortion cost is the smallest in consideration of the amount of bits and distortions generated when the current block is predicted and encoded for each of the frequency domain prediction weights and the candidate frequency domains transmitted from the candidate weighted frequency calculator. And a prediction frequency domain and a candidate frequency domain for each region.
The method of claim 6,

The frequency domain prediction unit,

Performing the color space prediction by subtracting a value obtained by applying a prediction weight of the selected frequency domain to a frequency coefficient of a color plane other than the reference color plane from a frequency coefficient of the selected frequency domain of the reference color plane of the transformed residual block. And a video encoding apparatus.
In the apparatus for decoding an image,

A decoder which decodes the encoded data to decode the color space prediction block;

Prediction weights for each frequency domain to be used for color space prediction based on the transform coefficients of blocks in the weighted candidate block set for each of the weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. And selecting the candidate frequency domain and using the prediction domain weights and the candidate frequency domains having the optimal prediction efficiency among the frequency domain prediction weights and the candidate frequency domains calculated for each of the N weight candidate blocksets. A color space reconstructor for reconstructing the residual block transformed from the predictive block;

An inverse transformer for inversely transforming the transformed residual block to restore the residual block;

A predictor for predicting a current block to generate a predicted block; And

An adder for reconstructing the current block by adding the reconstructed residual block and the prediction block;

Video decoding apparatus comprising a.
The method of claim 14,

The prediction weight for each frequency domain is

And an image decoding apparatus calculated by a correlation between the reference color plane and the remaining color planes.
The method of claim 14,

The prediction weight for each frequency domain is

An image decoding apparatus, characterized in that it is calculated in any one of a sequence unit, a frame unit, a macroblock unit and a subblock unit of an image.
The method of claim 14,

The color space restorer,

A frequency coefficient of the selected frequency domain of the reference color plane of the color space prediction block is added to a frequency coefficient of a color plane other than the reference color plane, and a value obtained by applying a prediction weight of the selected frequency domain to restore the converted residual block. And a video decoding apparatus.
The method of claim 14,

The color space restorer,

A weighted candidate selection unit for determining a set of N-weighted candidate blocks from the transform coefficient blocks of the decoded residual signal;

For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated based on a transform coefficient of a block in the weighting candidate block set, a prediction gain for each frequency domain is calculated from the prediction weights for the frequency domain, and the frequency domain is calculated. A candidate weighted frequency calculator configured to select a candidate frequency region to be used for color space prediction using the prediction gain for each star;

An optimum weighting filter determiner for receiving prediction weights and N candidate frequency domains for each of the N frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights and the candidate frequency domains for each frequency domain having an optimal prediction efficiency; And

A frequency domain reconstruction unit for reconstructing the residual block transformed from the color space prediction block by using the prediction weight for each frequency domain and the candidate frequency domain having the optimal prediction efficiency.

Video decoding apparatus comprising a.
The method of claim 18,

The N weight candidate blocksets

And a set of neighboring blocks and a decoded M (M is one or more natural numbers) macroblock row set.
The method of claim 18,

The N weight candidate blocksets

And a macroblock for each filter type of the macroblock belonging to the decoded macroblock row.
The method of claim 20,

The color space restorer,

A row prediction detector for checking whether color space prediction of one macroblock row is completed; And

After the color space prediction of the one macroblock row is completed, the weight of each type, which is a prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of macroblocks of each filter type of the one macroblock row. Filter type adjustment unit for resetting the filter type for the macroblock according to the type-specific prediction weights having the optimal filter efficiency for all macroblocks in the one macroblock row

The video decoding apparatus further comprises.
The method of claim 21,

The optimal filter rate is

And a rate-distortion cost is smallest in consideration of the amount of bits and distortions generated when the current block is predicted and encoded for each type-weighted value.
The method of claim 21,

The filter type adjustment unit,

And resetting the filter type if the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.
The method of claim 18,

The prediction weight for each frequency domain and the candidate frequency domain having the optimal prediction efficiency are

Frequency when the rate-distortion cost is the smallest in consideration of the amount of bits and distortions generated when the current block is predicted and encoded for each candidate frequency domain and the prediction weight for each frequency domain transmitted from the candidate weighted frequency calculator. An image decoding apparatus, characterized in that the prediction weight for each region and the candidate frequency domain.
The method of claim 18,

The frequency domain recovery unit,

A frequency coefficient of the selected frequency domain of the reference color plane of the color space prediction block is added to a frequency coefficient of a color plane other than the reference color plane, and a value obtained by applying a prediction weight of the selected frequency domain to restore the converted residual block. And a video decoding apparatus.
In the color space prediction apparatus of the transformed residual block,

A weighted candidate selection unit for determining a set of N-weighted candidate blocks from the transform coefficient blocks of the decoded residual signal;

For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated using a transform coefficient of a block in the weighted candidate block set, and a prediction gain for each frequency domain is calculated using the prediction weights for each frequency domain. A candidate weight frequency calculator for selecting a candidate frequency domain to be used for color space prediction by using the prediction gain for each frequency domain;

An optimum weighting filter determiner for receiving prediction weights and N candidate frequency domains for each of the N frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights and the candidate frequency domains for each frequency domain having an optimal prediction efficiency; And

A frequency domain prediction unit for generating a color space prediction block for the transformed residual block by using the prediction weights for each frequency domain and the candidate frequency domain having the optimal prediction efficiency.

Color space prediction apparatus comprising a.
The method of claim 26,

The N weight candidate blocksets are

And a set of neighboring blocks and a decoded M (M is one or more natural numbers) macroblock row set.
The method of claim 26,

The N weight candidate blocksets are

And a macroblock for each filter type of the macroblock belonging to the decoded macroblock row.
The method of claim 28,

The color space prediction device,

A row prediction detector for checking whether color space prediction of one macroblock row is completed; And

After the color space prediction of the one macroblock row is completed, the weight of each type, which is a prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of macroblocks of each filter type of the one macroblock row. Filter type adjustment unit for resetting the filter type for the macroblock according to the type-specific prediction weights having the optimal filter efficiency for all macroblocks in the one macroblock row

Color space prediction device further comprises.
The method of claim 29,

The optimal filter rate is

And a rate-distortion cost is the smallest in consideration of the amount of bits and distortions generated when the current block is predicted and encoded for each type weight.
The method of claim 29,

The filter type adjustment unit,

And resetting the filter type if the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.
The method of claim 26,

The prediction weight for each frequency domain is

And a color space prediction device calculated by a correlation between the reference color plane and the remaining color planes.
The method of claim 26,

The prediction weight for each frequency domain and the candidate frequency domain having the optimal prediction efficiency are

Frequency when the rate-distortion cost is the smallest in consideration of the amount of bits and distortions generated when the current block is predicted and encoded for each candidate frequency domain and the prediction weight for each frequency domain transmitted from the candidate weighted frequency calculator. Color space prediction apparatus, characterized in that the prediction weight for each region and the candidate frequency domain.
The method of claim 26,

The frequency domain prediction unit,

Performing the color space prediction by subtracting a value obtained by applying a prediction weight of the selected frequency domain to a frequency coefficient of a color plane other than the reference color plane from a frequency coefficient of the selected frequency domain of the reference color plane of the transformed residual block. Color space prediction device, characterized in that.
The method of claim 26,

The frequency domain prediction unit,

And performing color space prediction after quantizing the prediction weights of the selected frequency domain.
In the color space prediction apparatus of the color space prediction block,

A weighted candidate selection unit for determining a set of N-weighted candidate blocks from the transform coefficient blocks of the decoded residual signal;

For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated using a transform coefficient of a block in the weighted candidate block set, and a prediction gain for each frequency domain is calculated using the prediction weights for each frequency domain. A candidate weight frequency calculator for selecting a candidate frequency domain to be used for color space prediction by using the prediction gain for each frequency domain;

An optimum weighting filter determiner for receiving prediction weights and N candidate frequency domains for each of the N frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights and the candidate frequency domains for each frequency domain having an optimal prediction efficiency; And

A frequency domain reconstruction unit for reconstructing the residual block transformed from the color space prediction block by using the prediction weight for each frequency domain and the candidate frequency domain having the optimal prediction efficiency.

Color space prediction apparatus comprising a.
The method of claim 36,

The N weight candidate blocksets are

And a set of neighboring blocks and a decoded M (M is one or more natural numbers) macroblock row set.
The method of claim 36,

The N weight candidate blocksets are

And a macroblock for each filter type of the macroblock belonging to the decoded macroblock row.
The method of claim 38,

The color space prediction device,

A row prediction detector for checking whether color space prediction of one macroblock row is completed; And

After the color space prediction of the one macroblock row is completed, the weight of each type, which is a prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of macroblocks of each filter type of the one macroblock row. Filter type adjustment unit for resetting the filter type for the macroblock according to the type-specific prediction weights having the optimal filter efficiency for all macroblocks in the one macroblock row

Color space prediction device further comprises.
The method of claim 39,

The optimal filter rate is

And a rate-distortion cost is the smallest in consideration of the amount of bits and distortions generated when the current block is predicted and encoded for each type weight.
The method of claim 39,

The filter type adjustment unit,

And resetting the filter type if the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.
The method of claim 36,

The prediction weight for each frequency domain and the candidate frequency domain having the optimal prediction efficiency are

Frequency when the rate-distortion cost is the smallest in consideration of the amount of bits and distortions generated when the current block is predicted and encoded for each candidate frequency domain and the prediction weight for each frequency domain transmitted from the candidate weighted frequency calculator. Color space prediction apparatus, characterized in that the prediction weight for each region and the candidate frequency domain.
The method of claim 36,

The frequency domain recovery unit,

A frequency coefficient of the selected frequency domain of the reference color plane of the color space prediction block is added to a frequency coefficient of a color plane other than the reference color plane, and a value obtained by applying a prediction weight of the selected frequency domain to restore the converted residual block. Color space prediction device, characterized in that.
The method of claim 36,

The frequency domain recovery unit,

And performing color space prediction after quantizing the prediction weights of the selected frequency domain.
In the method of encoding / decoding an image,

A prediction block is generated by predicting a current block for each color plane, and a residual block is generated by subtracting the prediction block from the current block, the residual block is transformed, and a selected N (from a transform coefficient block of a decoded residual signal). N is a weighted candidate block set for each frequency domain to be used for color space prediction based on the transform coefficients of the blocks in the weighted candidate block set for each of the weighted candidate block sets, and the N weighted candidate blocks. A color space prediction block is generated for the transformed residual block by using the prediction weighted frequency and the candidate frequency domain having an optimal prediction efficiency among the frequency domain prediction weights and the candidate frequency domains calculated for each set. Encoding a color space prediction block; And

Decoding the coded data to decode the color space prediction block, and for each N weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal, transform coefficients of the blocks in the weighted candidate block set. Select a prediction weight region and a candidate frequency region for each frequency domain to be used for color space prediction based on the frequency domain having an optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weight candidate blocksets Reconstruct the residual block transformed from the color space prediction block by using the prediction weights for each candidate and the candidate frequency domain, restore the residual block by inversely transforming the transformed residual block, generate a prediction block by predicting a current block, and Adding the reconstructed residual block and the prediction block to the Steps to Restore Current Block

Image encoding / decoding method comprising a.
In the method of encoding an image,

Generating a prediction block by predicting the current block for each color plane;

Subtracting the prediction block from the current block to generate a residual block;

Transforming the residual block;

Prediction weights for each frequency domain to be used for color space prediction based on the transform coefficients of blocks in the weighted candidate block set for each of the weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. And selecting the candidate frequency domain and converting the prediction weighted frequency and the candidate frequency domain by the frequency domain with the optimal prediction efficiency among the frequency domain prediction weights and candidate frequency domains calculated for each of the N weighted candidate blocksets. A color space prediction step of generating a color space prediction block for the residual block; And

Encoding the color space prediction block

Image encoding method comprising a.
The method of claim 46,

The color space prediction step,

The color space prediction block is generated by subtracting a value obtained by applying a prediction weight of the selected frequency domain to a frequency coefficient of a color plane other than the reference color plane from a frequency coefficient of the selected frequency domain of the reference color plane of the transformed residual block. And a video encoding method.
The method of claim 46,

The color space prediction step,

Determining a set of N weighted candidate blocks from the transform coefficient block of the decoded residual signal;

For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated based on the transform coefficients of the blocks in the weighting candidate block set, a prediction gain for each frequency domain is calculated from the prediction weights for the frequency domain, and the frequency domain is calculated. Selecting a candidate frequency region to be used for color space prediction by using the predictive gain for each star;

Setting candidate frequency domains and weights having an optimal prediction efficiency among candidate frequency domains and weights calculated for each of the N weight candidate blocks, as prediction frequency domains and weights; And

Frequency domain prediction step of generating a color space prediction block for the transformed residual block from the prediction weight of the selected frequency domain

Image encoding method comprising a.
The method of claim 48,

The N weight candidate blocksets are

And a macroblock for each filter type of the macroblock belonging to the decoded macroblock row.
The method of claim 49,

The color space prediction step,

A row prediction detection step of checking whether color space prediction of one macroblock row is completed; And

After the color space prediction of the one macroblock row is completed, the weight of each type, which is a prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of macroblocks of each filter type of the one macroblock row. A filter type adjustment step of resetting a filter type for the macroblock according to a type-specific prediction weight having an optimal filter efficiency for all macroblocks in the one macroblock row;

The image encoding method further comprises.
51. The method of claim 50,

In the filter type adjustment step,

And resetting the filter type if the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.
The method of claim 48,

The frequency domain frequency domain prediction step,

Performing the color space prediction by subtracting a value obtained by applying a prediction weight of the selected frequency domain to a frequency coefficient of a color plane other than the reference color plane from a frequency coefficient of the selected frequency domain of the reference color plane of the transformed residual block. Video encoding method.
In the method of decoding an image,

Decoding the color space prediction block by decoding the encoded data;

Prediction weights for each frequency domain to be used for color space prediction based on the transform coefficients of blocks in the weighted candidate block set for each of the weighted candidate block sets selected from the transform coefficient blocks of the decoded residual signal. And selecting the candidate frequency domain and using the prediction domain weights and the candidate frequency domains having the optimal prediction efficiency among the frequency domain prediction weights and the candidate frequency domains calculated for each of the N weight candidate blocksets. A color space reconstruction step of reconstructing the residual block transformed from the prediction block;

An inverse transform step of inversely transforming the transformed residual block to restore a residual block;

Predicting a current block to generate a predicted block; And

Reconstructing the current block by adding the reconstructed residual block and the prediction block

Image decoding method comprising a.
The method of claim 53,

The color space restoration step,

A frequency coefficient of the selected frequency domain of the reference color plane of the color space prediction block is added to a frequency coefficient of a color plane other than the reference color plane, and a value obtained by applying a prediction weight of the selected frequency domain to restore the converted residual block. Video decoding method characterized in that the.
The method of claim 53,

The color space restoration step,

Determining a set of N weighted candidate blocks from the transform coefficient block of the decoded residual signal;

For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated based on the transform coefficients of the blocks in the weighting candidate block set, a prediction gain for each frequency domain is calculated from the prediction weights for the frequency domain, and the frequency domain is calculated. Selecting a candidate frequency region to be used for color space prediction by using the predictive gain for each star;

Setting candidate frequency domains and weights having an optimal prediction efficiency among candidate frequency domains and weights calculated for each of the N weight candidate blocks, as prediction frequency domains and weights; And

A frequency domain restoration step of restoring a residual block transformed from the color space prediction block by using the prediction weight of the selected frequency domain

Image decoding method comprising a.
The method of claim 55,

The N weight candidate blocksets are

A set of neighboring blocks and a set of decoded M (M is one or more natural numbers) macroblock row sets.
The method of claim 55,

The N weight candidate blocksets are

And a macroblock for each filter type of the macroblock belonging to the decoded macroblock row.
The method of claim 57,

The color space restoration step,

A row prediction detection step of checking whether color space prediction of one macroblock row is completed; And

After the color space prediction of the one macroblock row is completed, the weight of each type, which is a prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of macroblocks of each filter type of the one macroblock row. A filter type adjustment step of resetting a filter type for the macroblock according to a type-specific prediction weight having an optimal filter efficiency for all macroblocks in the one macroblock row;

The image decoding method further comprises.
The method of claim 58,

In the filter type adjustment step,

And resetting the filter type if the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.
In the color space prediction method of the transformed residual block,

Determining a set of N weighted candidate blocks from the transform coefficient block of the decoded residual signal;

For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated based on the transform coefficients of the blocks in the weighting candidate block set, a prediction gain for each frequency domain is calculated from the prediction weights for the frequency domain, and the frequency domain is calculated. Selecting a candidate frequency region to be used for color space prediction by using the predictive gain for each star;

Receiving prediction weights for each N frequency domains and N candidate frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights for each frequency domain and the candidate frequency domains having optimal prediction efficiency; And

A frequency domain prediction step of generating a color space prediction block for the transformed residual block by using the prediction weights for each frequency domain and the candidate frequency domain having the optimal prediction efficiency.

Color space prediction method comprising a.
The method of claim 60,

The N weight candidate blocksets

And a macroblock for each filter type of the macroblock belonging to the decoded macroblock row.
62. The method of claim 61,

The color space prediction step,

A row prediction detection step of checking whether color space prediction of one macroblock row is completed; And

After the color space prediction of the one macroblock row is completed, the weight of each type, which is a prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of macroblocks of each filter type of the one macroblock row. A filter type adjustment step of resetting a filter type for the macroblock according to a type-specific prediction weight having an optimal filter efficiency for all macroblocks in the one macroblock row;

The color space prediction method further comprises.
The method of claim 62,

In the filter type adjustment step,

And resetting the filter type if the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.
In the color space prediction method of the color space prediction block,

Determining a set of N weighted candidate blocks from the transform coefficient block of the decoded residual signal;

For each of the N weighted candidate blocksets, a prediction weight for each frequency domain is calculated based on the transform coefficients of the blocks in the weighting candidate block set, a prediction gain for each frequency domain is calculated from the prediction weights for the frequency domain, and the frequency domain is calculated. Selecting a candidate frequency region to be used for color space prediction by using the predictive gain for each star;

Receiving prediction weights for each N frequency domains and N candidate frequency domains, calculating and comparing the prediction efficiency for each of the frequency domains, and determining the prediction weights for each frequency domain and the candidate frequency domains having optimal prediction efficiency; And

A frequency domain reconstruction step of reconstructing the residual block transformed from the color space prediction block using the prediction domain weight and the candidate frequency domain for each frequency domain having the optimal prediction efficiency.

Color space prediction method comprising a.
The method of claim 64, wherein

The N weight candidate blocksets are

And a macroblock for each filter type of the macroblock belonging to the decoded macroblock row.
66. The method of claim 65,

The color space prediction step,

A row prediction detection step of checking whether color space prediction of one macroblock row is completed; And

After the color space prediction of the one macroblock row is completed, the weight of each type, which is a prediction weight for each frequency domain, is calculated based on the transform coefficient of the macroblock for each set of macroblocks of each filter type of the one macroblock row. A filter type adjustment step of resetting a filter type for the macroblock according to a type-specific prediction weight having an optimal filter efficiency for all macroblocks in the one macroblock row;

The color space prediction method further comprises.
The method of claim 66,

The filter type adjustment step,

And resetting the filter type if the number of the reset filter types is greater than or equal to a preset number as a result of resetting the filter type.