KR101464944B1 - Video Encoding/Decoding Method and Apparatus Using Arbitrary Pixel in Subblock - Google Patents

Abstract

The present invention relates to a method and apparatus for image encoding / decoding using arbitrary pixels in a sub-block.
The present invention relates to an arbitrary pixel extracting unit. A first predictor; A first subtractor; A first conversion unit; A first quantization unit; A first inverse quantization unit; A first inverse transform unit; A first adder; A second predictor; A second subtraction unit; A random residual pixel combiner; A second conversion unit; A second quantization unit; And an encoding unit. The encoding unit extracts, encodes, and decodes arbitrary pixels in each sub-block of the current macroblock, reconstructs and restores the rest of the sub-blocks, and predicts and codes the remaining pixels of each sub-block.
According to the present invention, in coding or decoding an image, it is possible to improve coding efficiency by improving the accuracy of prediction when intraprediction of a pixel of a block to be coded or decoded at present is performed.

Description

[0001] The present invention relates to a video encoding / decoding method and apparatus using arbitrary pixels in a sub-block,

The present invention relates to a method and apparatus for image encoding / decoding using arbitrary pixels in a sub-block. More particularly, the present invention relates to a method and an apparatus for improving coding efficiency by improving the accuracy of prediction when intra-prediction is performed on a pixel of a current block to be coded or decoded in coding or decoding an image.

Moving Picture Experts Group (MPEG) and Video Coding Experts Group (VCEG) have developed superior video compression techniques superior to the existing MPEG-4 Part 2 and H.263 standards. This new standard is called H.264 / AVC (Advanced Video Coding) and is jointly announced as MPEG-4 Part 10 AVC and ITU-T Recommendation H.264. In this H.264 / AVC (abbreviated as 'H.264'), a spatial prediction coding method which is different from the international standard related to the conventional video coding such as MPEG-1, MPEG-2 and MPEG- Predictive Coding) method.

In the conventional moving image encoding method, "intra prediction" is used for the coefficient value transformed in the discrete cosine transform domain (DCT domain), thereby seeking to increase the coding efficiency. As a result, Resulting in deterioration in the subjective image quality of the transmission bit rate band. However, H.264 adopts a coding method based on Spatial Intra Prediction in a Spatial Domain instead of a Transform Domain.

An encoder using an encoding method based on a conventional spatial intra prediction predicts a current block to be coded from information of a previous block that has been already encoded and reproduced and extracts a difference from an actual block to be encoded, Only information is encoded and transmitted to a decoder. At this time, parameters necessary for predicting the block may be transmitted to the decoder, or the encoder and the decoder may be synchronized to share parameters necessary for prediction, so that the decoder may predict the block. The decoder predicts the current block using the decoded neighboring block and generates a sum of the difference information and the predicted current block transmitted from the encoder to generate a desired block to be decoded. At this time, if parameters necessary for prediction are also transmitted from the encoder, the decoder uses the parameter to predict the current block.

However, in the case of predicting using intra prediction according to a conventional image coding method or an image decoding method, it is possible to predict a previously reconstructed adjacent pixel in a neighboring block (mainly a left block and an upper block of the current block) To predict the pixels of the current block. In this case, since the pixels of the current block located far away from the neighboring pixels of the neighboring block are predicted as the pixel values of the neighboring pixels farther away, the prediction accuracy may be lowered. This decrease in the prediction accuracy increases the difference between the pixel value of the original pixel and the pixel value of the predicted pixel, which ultimately causes a problem of lowering the compression efficiency.

In order to solve the above-described problems, the present invention has a main purpose in improving coding efficiency by improving the accuracy of prediction when intra-prediction of a pixel of a block to be coded or decoded is performed in coding or decoding an image.

According to an aspect of the present invention, there is provided an apparatus for encoding an image, comprising: an arbitrary pixel extracting unit for extracting arbitrary pixels from each sub-block of a current macroblock of an image to generate an arbitrary pixel block; A first predictor for generating a predictive arbitrary pixel block by predicting an arbitrary pixel block; A first subtractor for generating an arbitrary pixel residual block by subtracting the arbitrary pixel block and the predictive arbitrary pixel block; A first conversion unit for converting an arbitrary pixel residual block; A first quantization unit for quantizing the transformed arbitrary-pixel residual block; A first inverse quantization unit for inversely quantizing the quantized arbitrary pixel residual block; A first inverse transformer for inversely transforming the inversely quantized arbitrary pixel residual block; A first adder for adding the predictive arbitrary pixel block and the inverse-transformed arbitrary pixel residual block to generate a reconstructed arbitrary pixel block; A second predictor for generating a predictive macroblock by predicting remaining pixels of each sub-block using adjacent pixels of the current macroblock and reconstructed arbitrary pixels of the reconstructed arbitrary pixel block; A second subtractor for subtracting the predictive macroblock from the current macroblock to generate a residual macroblock; An arbitrary residual pixel combining unit which combines each inverse transformed residual pixel of the inversely transformed arbitrary pixel residual block to each position of the residual macroblock; A second transform unit for transforming the residual macroblock to which each inverse transformed residual pixel is combined; A second quantization unit for quantizing the transformed residual macroblock; And a coding unit for coding the quantized residual macroblock to generate a bitstream.

According to another aspect of the present invention, there is provided an apparatus for encoding an image, comprising: an arbitrary pixel extracting unit for extracting arbitrary pixels in each sub-block of a current macroblock of an image to generate an arbitrary pixel block; A first predictor for generating a predictive arbitrary pixel block by predicting an arbitrary pixel block; A first subtractor for generating an arbitrary pixel residual block by subtracting the arbitrary pixel block and the predictive arbitrary pixel block; A first conversion unit for converting an arbitrary pixel residual block; A first quantization unit for quantizing the transformed arbitrary-pixel residual block; A first inverse quantization unit for inversely quantizing the quantized arbitrary pixel residual block; A first inverse transformer for inversely transforming the inversely quantized arbitrary pixel residual block; A first adder for adding the predictive arbitrary pixel block and the inverse-transformed arbitrary pixel residual block to generate a reconstructed arbitrary pixel block; A second predictor for generating each predicted sub-block by predicting remaining pixels of each sub-block using adjacent pixels of each sub-block and each reconstructed arbitrary pixel of the reconstructed arbitrary pixel block; A second subtractor for subtracting each predicted sub-block from each sub-block to generate each residual sub-block; An arbitrary residual pixel combiner for combining each inverse transformed residual pixel of the inversely transformed arbitrary pixel residual block to each position of each residual sub-block; A second transform unit for transforming each residual sub-block to which each inverse transformed residual pixel is combined; A second quantization unit for quantizing each transformed residual sub-block; And a coding unit for coding each quantized residual sub-block to generate a bitstream, wherein each sub-block is coded and decoded, and then the next sub-block is coded and decoded. to provide.

According to another aspect of the present invention, there is provided a method of encoding an image, comprising: an arbitrary pixel extracting step of extracting arbitrary pixels in each sub-block of a current macroblock of an image to generate an arbitrary pixel block; A first prediction step of generating a prediction arbitrary pixel block by predicting an arbitrary pixel block; A first subtraction step of generating an arbitrary pixel residual block by subtracting an arbitrary pixel block and a predictive arbitrary pixel block; A first conversion step of converting an arbitrary pixel residual block; A first quantization step of quantizing the transformed arbitrary-pixel residual block; A first inverse quantization step of inversely quantizing the quantized arbitrary pixel residual block; A first inverse transformation step of inversely transforming the inversely quantized arbitrary pixel residual block; A first adding step of adding a predictive arbitrary pixel block and an inverse-transformed arbitrary pixel residual block to generate a reconstructed arbitrary pixel block; A second prediction step of generating a predictive macroblock by predicting remaining pixels of each sub-block using adjacent pixels of the current macroblock and reconstructed arbitrary pixels of the reconstructed arbitrary pixel block; A second subtraction step of subtracting a predictive macroblock from a current macroblock to generate a residual macroblock; An arbitrary residual pixel combining step of combining each inverse transformed residual pixel of the inverse transformed arbitrary pixel residual block to each position of the residual macroblock; A second conversion step of converting a residual macroblock to which each inverse transformed residual pixel is combined; A second quantization step of quantizing the transformed residual macroblock; And a coding step of coding the quantized residual MK block to generate a bitstream.

According to still another aspect of the present invention, there is provided a method of encoding an image, comprising: an arbitrary pixel extracting step of extracting arbitrary pixels in each sub-block of a current macroblock of an image to generate an arbitrary pixel block; A first prediction step of generating a prediction arbitrary pixel block by predicting an arbitrary pixel block; A first subtraction step of generating an arbitrary pixel residual block by subtracting an arbitrary pixel block and a predictive arbitrary pixel block; A first conversion step of converting an arbitrary pixel residual block; A first quantization step of quantizing the transformed arbitrary-pixel residual block; A first inverse quantization step of inversely quantizing the quantized arbitrary pixel residual block; A first inverse transformation step of inversely transforming the inversely quantized arbitrary pixel residual block; A first adding step of adding a predictive arbitrary pixel block and an inverse-transformed arbitrary pixel residual block to generate a reconstructed arbitrary pixel block; A second prediction step of generating each predicted sub-block by predicting remaining pixels of each sub-block using adjacent pixels of each sub-block and each reconstructed arbitrary pixel of the reconstructed arbitrary pixel block; A second subtraction step of subtracting each predicted sub-block from each sub-block to generate each residual sub-block; An arbitrary residual pixel combining step of combining each inverse transformed residual pixel of the inversely transformed arbitrary pixel residual block to each position of each residual sub-block; A second transforming step of transforming each residual sub-block to which each inverse transformed residual pixel is combined; A second quantization step of quantizing each transformed residual sub-block; And a coding step of coding each quantized residual sub-block to generate a bitstream, wherein each sub-block is coded and decoded, and the next sub-block is coded and decoded.

According to another aspect of the present invention, there is provided an apparatus for decoding an image, the apparatus comprising: a decoding unit decoding a bitstream to extract a residual macroblock and a prediction mode; A dequantizer for dequantizing the residual macroblock; A coefficient extraction unit for extracting a frequency coefficient corresponding to a position of an arbitrary pixel of each sub-block in an inversely quantized residual macroblock; A first inverse transformer for inversely transforming the extracted frequency coefficients to generate an inverse transformed arbitrary pixel residual block; A first predictor for generating a predictive arbitrary pixel block by predicting an arbitrary pixel block according to a prediction direction according to a prediction mode; A first adder for adding the predictive arbitrary pixel block and the inverse-transformed arbitrary pixel residual block to generate a reconstructed arbitrary pixel block; A second inverse transformer for inversely transforming the dequantized residual macroblock to generate an inverse transformed residual macroblock; A second predictor for generating a predictive current macroblock by predicting a current macroblock of an image according to a prediction direction according to a prediction mode using adjacent pixels of the current macroblock of the image and reconstructed arbitrary pixels of the reconstructed arbitrary pixel block, ; A second adder for generating a reconstructed current macroblock using the predictive current macroblock and the inverse transformed residual macroblock; And an arbitrary pixel combining unit for combining the reconstructed arbitrary pixel with each position of the reconstructed current macroblock.

According to another aspect of the present invention, there is provided a method of decoding an image, the decoding method comprising: a decoding step of decoding a bitstream to extract a residual macroblock and a prediction mode; An inverse quantization step of dequantizing the residual macroblock; A coefficient extracting step of extracting a frequency coefficient corresponding to a position of an arbitrary pixel of each sub-block in an inversely quantized residual macroblock; A first inverse transform step of inversely transforming the extracted frequency coefficients to generate a reconstructed arbitrary pixel residual block; A first prediction step of generating a prediction arbitrary pixel block by predicting an arbitrary pixel block according to a prediction direction according to a prediction mode; A first addition step of generating an arbitrary pixel block reconstructed by adding a predictive arbitrary pixel block and a reconstructed arbitrary pixel residual block; A second inverse transform step of inversely transforming the inversely quantized residual macroblock to generate an inverse transformed residual macroblock; A second prediction step of generating a prediction current macroblock by predicting a current macroblock of an image according to a prediction direction according to a prediction mode using adjacent pixels of the current macroblock of the image and reconstructed arbitrary pixels of the reconstructed arbitrary pixel block, ; A second addition step of generating a reconstructed current macroblock using the prediction current macroblock and the inverse transformed residual macroblock; And an arbitrary pixel combining step of combining each reconstructed arbitrary pixel with each position of the reconstructed current macroblock.

As described above, according to the present invention, in coding or decoding an image, when the pixels of a current block to be coded or decoded are intra-predicted, the accuracy of prediction can be improved to improve the coding efficiency.

1 is a block diagram schematically showing an electronic configuration of a video encoding apparatus;
FIGS. 2 and 3 are diagrams illustrating an intra-prediction mode,
Figs. 4 and 5 are diagrams showing an example of a basic vector for DCT and inverse transform,
6 is a diagram illustrating an arbitrary pixel in a sub-block according to an exemplary embodiment of the present invention.
FIG. 7 is a block diagram schematically illustrating an electronic configuration of an image encoding apparatus according to an embodiment of the present invention; FIG.
FIG. 8 is an exemplary view showing an arbitrary pixel block extracted from a current macroblock,
FIG. 9 is a diagram illustrating an example of predicting an arbitrary pixel block using adjacent pixels of a current macroblock;
Figs. 10 and 11 are examples showing modified basic vectors for DCT and inverse transform,
FIGS. 12 to 15 are diagrams illustrating examples of predicting remaining pixels of each sub-block,
16 is an exemplary diagram showing one-dimensional transform for two-dimensional DCT transform,
17 and 18 are views illustrating a process of inverse conversion using a modified basic vector according to the order of the inverse conversion directions,
FIG. 19 is a flowchart illustrating an image encoding method according to an embodiment of the present invention. FIG.
20 is a flowchart illustrating a process of restoring a current macroblock according to an image encoding method according to an embodiment of the present invention.
FIG. 21 is a block diagram schematically illustrating an electronic configuration of an image decoding apparatus according to an embodiment of the present invention; FIG.
FIG. 22 is a flowchart illustrating a video decoding method according to an embodiment of the present invention. FIG.
FIG. 23 is a diagram illustrating a procedure for coding each sub-block according to another embodiment of the present invention; FIG.
FIG. 24 is a diagram illustrating an example of determining a prediction direction of each sub-block according to another embodiment of the present invention; FIG.
FIG. 25 is a diagram showing an encoding / decoding procedure of each sub-block according to a raster scan direction;
FIG. 26 is an exemplary view showing a state in which each sub-block is predicted according to a newly determined optimum prediction direction;
FIGS. 27 and 28 illustrate a process of predicting remaining frequency coefficients using a part of frequency coefficients according to an embodiment of the present invention and another embodiment.

Hereinafter, some embodiments of the present invention will be described in detail with reference to exemplary drawings. It should be noted that, in adding reference numerals to the constituent elements of the drawings, the same constituent elements are denoted by the same reference numerals whenever possible, even if they are shown in different drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

In describing the components of the present invention, terms such as first, second, A, B, (a), and (b) may be used. These terms are intended to distinguish the constituent elements from other constituent elements, and the terms do not limit the nature, order or order of the constituent elements. When a component is described as being "connected", "coupled", or "connected" to another component, the component may be directly connected or connected to the other component, Quot; may be "connected," "coupled," or "connected. &Quot;

1 is a block diagram schematically showing an electronic configuration of a video encoding apparatus.

The image encoding apparatus 100 includes a predictor 110, a subtractor 120, a transformer 130, a quantizer 140, a coder 150, an inverse quantizer 160, An inverse transform unit 170, and an adder 180. [

The prediction unit 110 generates a prediction block by predicting a current block to be coded in the image. That is, the predicting unit 110 predicts the pixel value of each pixel of the current block to be encoded in the image according to a predetermined optimal prediction mode and outputs the predicted pixel value Predicted Pixel Value) is generated. The prediction unit 110 may transmit information on the prediction mode to the encoding unit 150 so that the encoding unit 150 may encode information on the prediction mode.

Here, the optimal prediction mode is one of various intra prediction modes for intra prediction (i.e., nine prediction modes in intra 8x8 prediction and intra 4x4 prediction, and four prediction modes in intra 16x16 prediction) The prediction mode in which the coding cost is the minimum can be determined.

FIGS. 2 and 3 are exemplary diagrams showing an intra-prediction mode.

2, the intra 4x4 prediction includes a vertical mode, a horizontal mode, a DC (direct current) mode, a diagonal down-left mode, a diagonal down- There are nine prediction modes including a vertical-right mode, a horizontal-down mode, a vertical-left and a horizontal-up mode. Intra 8x8 prediction There are nine prediction modes similar to intra 4x4 prediction. Also, referring to FIG. 3, intra-16 × 16 prediction has four prediction modes including a vertical mode, a horizontal mode, a DC mode, and a plane mode.

The prediction unit 110 may calculate the coding cost according to the nine prediction modes or the four prediction modes shown in the block mode or the block size of the current block and determine the prediction mode having the minimum coding cost as the optimal prediction mode .

Subtraction unit 120 subtracts the prediction block from the current block to generate a residual block. That is, the subtractor 120 calculates the 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 predicted block predicted by the predictor 110, and outputs a residual signal ). ≪ / RTI >

The transforming unit 130 transforms the residual block into the frequency domain and transforms each pixel value of the residual block into a frequency coefficient. Here, the transform unit 130 may transform the image signal on the time axis, such as a Hadamard transform, a discrete cosine transform based transform (DCT) transform, The residual signal can be transformed into the frequency domain, and the residual signal transformed into the frequency domain becomes the frequency coefficient.

FIGS. 4 and 5 illustrate exemplary basic vectors for DCT and inverse transform. For example, the transforming unit 130 performs a one-dimensional DCT in the horizontal direction and a one-dimensional DCT in the vertical direction using the basic vector shown in FIG. 4 created by Equation (1) Thereby converting the pixel value of the spatial domain into the frequency domain. At this time, when the block mode is intra 16x16, only the frequency coefficients of 16 (4x4) DC components are collected for each of 16 sub-blocks of 4x4 block size, and then Hadamard transform is performed as shown in Equation (2).

The quantization unit 140 quantizes the residual block having the residual signal transformed into the frequency domain by the transform unit 130. The quantization unit 240 quantizes the transformed residual block using a dead zone uniform threshold quantization (DZUTQ), a quantization weight matrix or a quantization technique such as a modified quantization weight matrix Can be quantized.

The encoding unit 150 outputs the bitstream by encoding the transformed and quantized residual block using an entropy coding technique or the like. At this time, the encoder 150 scans the quantized frequency coefficients of the quantized residual block by various methods to generate a quantized frequency coefficient string, and outputs the quantized frequency coefficient string as a bit stream using an entropy coding method or the like . Also, the encoding unit 150 may encode information on a prediction mode in which the current block is predicted by the predicting unit 110 together.

The inverse quantization unit 160 performs inverse quantization on the residual blocks quantized by the quantization unit 140. That is, the dequantizer 160 dequantizes the quantization frequency coefficients of the coarse residual block to generate a residual block having a frequency coefficient.

The inverse transform unit 170 performs an inverse transform on the inversely quantized residual block by the inverse quantization unit 160. That is, the inverse transform unit 170 inversely transforms the frequency coefficients of the inversely quantized residual block to generate a residual block having a pixel value, that is, a reconstructed residual block. Here, the inverse transform unit 170 may inversely transform the transformed transformed transformed image using the transformed transformed transformed image. For example, the inverse transform unit 170 collects the frequency coefficients of the DC component, performs inverse Hadamard transformation as shown in Equation (3), and uses the basic vector shown in FIG. 5 to transform the current block into a one- DCT and performs inverse one-dimensional DCT in the horizontal direction to inversely transform the frequency coefficient of the frequency domain into the spatial domain pixel value.

The adder 180 adds the prediction block predicted by the predictor 110 and the residual block reconstructed by the inverse transformer 170 to reconstruct the current block. The reconstructed current block is stored in the predictor 110 as a reference picture, and is used as a reference picture when coding a next block or a future block of the current block.

Although not shown in FIG. 1, a deblocking filter unit (not shown) may be additionally connected between the predictor 110 and the adder 180. The deblocking filter unit deblocking filters the current block restored by the adder 180. Here, deblocking filtering refers to a task of reducing block distortion caused by coding an image block by block. The deblocking filter may be applied to a block boundary or a macro block boundary, a deblocking filter may be applied to a macro block boundary, Can be selectively used.

FIG. 6 is a diagram illustrating an arbitrary pixel in a sub-block according to an exemplary embodiment of the present invention. Referring to FIG.

In encoding a macroblock of an arbitrary size, an arbitrary pixel, which is a pixel at an arbitrary position in a sub-block, is encoded in units of sub-blocks of an arbitrary size, Predicts the remaining pixels in the sub-block so that all the pixels in the macro-block can be predicted from nearby pixels instead of being predicted from neighboring pixels at distant positions, thereby preventing the accuracy of prediction from being lowered.

Here, the macroblock 610 of an arbitrary size may be designated by the user as the size of NxM, and any size of the subblock 620 of any size within the macroblock 610 of arbitrary size may be specified by the user The size may be designated as the OxP size (where O? N, P? M).

FIG. 6 shows an arbitrary pixel 630 when a macroblock 610 of an arbitrary size to be encoded is a 16x16 block and a sub-block 620 is a 4x4 block. In the following examples, the present invention will be described by way of examples. In the present invention, any pixel is determined as a pixel located at the right end and the bottom end in a sub-block and is referred to as X _i (630). However, the size of the macro block, the size of the sub-block, the position of the arbitrary pixel, and the like are illustrative, and pixels of various block sizes or various positions can be used.

FIG. 7 is a block diagram schematically illustrating an electronic configuration of an image encoding apparatus according to an embodiment of the present invention. Referring to FIG.

An image encoding apparatus 700 according to an embodiment of the present invention encodes an image and includes an arbitrary pixel extracting unit 702, a first predicting unit 704, a first subtracting unit 706, A second inverse quantization unit 708, a first quantization unit 710, a first inverse quantization unit 712, a first inverse transformation unit 714, a first addition unit 716, a first prediction unit 718, A second inverse quantization unit 720, an arbitrary residual pixel combination unit 7220, a second transform unit 724, a second quantization unit 726, an encoding unit 728, a second inverse quantization unit 730, 732, a second adder 734, and an arbitrary reconstructed pixel combiner 736.

The image encoding apparatus 700 may include a personal computer (PC), a notebook computer, a personal digital assistant (PDA), a portable multimedia player (PMP) A PlayStation Portable (PSP), a Mobile Communication Terminal, or the like, and may be a communication device such as a communication modem for performing communication with various devices or wired / wireless communication networks, various programs and data for encoding an image A memory for storing the program, and a microprocessor for executing and calculating the program and controlling the program.

The arbitrary-pixel extracting unit 702 extracts arbitrary pixels in each sub-block of the current macroblock of the image to generate an arbitrary pixel block. That is, the arbitrary-pixel extracting unit 700 extracts arbitrary pixels 630 X ₀ to X ₁₅ at arbitrary positions among the pixels in each sub-block 620 of the current macroblock 610 shown in FIG. 6 And an arbitrary pixel block is generated as shown in Fig. FIG. 8 illustrates an example of extracting an arbitrary pixel block from a current macroblock.

The first predictor 704 predicts an arbitrary pixel block to generate a predictive arbitrary pixel block. That is, the first predicting unit 704 is adjacent to the current macroblock among the pixels of the left and upper macroblocks of the current macroblock as illustrated in FIG. 8, and before the current macroblock is coded using the pixel _{V i (V 0 ~ V 15} ) and _{H i (H 0 ~ H 15} ) has already been decoded as the adjacent pixels for prediction of any pixel block, a predicted any predicts each arbitrary pixels of any pixel block Thereby generating a predictive arbitrary pixel block having pixels.

In addition, the first predictor 704 may perform low-pass filtering on neighboring pixels of the current macroblock and predict an arbitrary pixel block using the downsampled pixels. FIG. 9 illustrates an example of predicting an arbitrary pixel block using adjacent pixels of a current macroblock. 9, the first predictor 704 low-pass-filters the V _i (V ₀ to V ₁₅ ) among the neighboring pixels of the current macroblock and downsamples the result to obtain V _a , V _b , V _c , and V _d The left adjacent pixel of the arbitrary pixel block can be determined. Similarly, the first prediction unit 704 is an optionally by calculating the current of the adjacent pixels of the macro-block _{H i (H 0 ~ H 15} ) a low pass to downsampling after filtering H _a, H _b, H _c, H _d The upper adjacent pixel of the pixel block can be determined.

At this time, the first predictor 704 can predict an arbitrary pixel block in the same prediction direction as the prediction direction according to the prediction mode of the current macroblock. Here, the prediction direction of the current macroblock may be a prediction direction in accordance with the prediction mode of the intra 16x16 prediction as shown in FIG. 3, and it may be predicted that prediction cost according to each prediction mode is minimized Direction as an optimal prediction direction and can transmit information on the prediction mode to the encoding unit 728. [

The first subtractor 706 subtracts the arbitrary pixel block and the predictive arbitrary pixel block to generate an arbitrary pixel residual block. That is, the first subtractor 706 subtracts an arbitrary pixel residual block, which is a residual block having a residual signal having a value obtained by subtracting the pixel value of each arbitrary pixel in the arbitrary pixel block from the pixel value of each predictive arbitrary pixel in the predictive arbitrary pixel block .

The first transform unit 708 transforms the arbitrary pixel residual block. That is, the first converter 708 converts the arbitrary residual signals of the arbitrary pixel residual block into the frequency domain to generate the frequency coefficients for the arbitrary residual pixels. Also, the first conversion unit 708 can independently convert only the arbitrary pixel residual signals of the arbitrary pixel residual block. To this end, the first transform unit 708 can perform DCT transform on each arbitrary-pixel residual signal of the arbitrary-pixel residual block using the modified basic vector for DCT transform, thereby independently transforming only the arbitrary-pixel residual signal .

Also, the first transform unit 708 may perform the Hadamard transform after performing the DCT transform. That is, when the current macroblock is an intra 16x16 block, the first transform unit 708 can perform Hadamard transform on the DCT-transformed frequency coefficients. If the current macroblock is an intra 4x4 block, Hadamard transform is added . ≪ / RTI > If the current macroblock is determined to be an intra 4x4 block, the efficiency of the Hadamard transformation may be lowered because the correlation of the image is small.

The first quantization unit 710 quantizes the converted arbitrary pixel residual block. That is, the first quantization unit 710 quantizes the frequency coefficients of the arbitrary pixel residual block to generate an arbitrary pixel residual block having the quantization frequency coefficient.

The first dequantization unit 710 dequantizes the quantized arbitrary pixel residual block. That is, the first dequantization unit 710 dequantizes the quantization frequency coefficients of the arbitrary pixel residual block to generate an arbitrary pixel residual block having a frequency coefficient.

The first inverse transform unit 712 inversely transforms the inversely quantized arbitrary pixel residual block. That is, the first inverse transform unit 712 transforms the frequency coefficients of the arbitrary pixel residual block into a spatial domain, and generates a reconstructed arbitrary pixel residual block having the reconstructed arbitrary residual signal. At this time, the first inverse transform unit 712 performs inverse discrete cosine transform using the modified basic vectors for DCT inverse transform of the frequency coefficients in which the arbitrary pixel residual signals are transformed, Can be converted. In addition, the first inverse transform unit 712 can perform inverse discrete cosine transform after inverse Hadamard transform of the transformed frequency coefficients of the arbitrary pixel residual signals.

Here, the modified basic vector for the DCT transform and the DCT inverse transform is a method of transforming an arbitrary pixel independently of the remaining pixels of each subblock and inverse transforming the DCT transform basic vector and the DCT transform A conversion basic vector, that is, a basic vector obtained by improving existing basic vectors, can be determined as shown in FIGS. 10 and 11. FIG. 10 and 11 illustrate modified basic vectors for DCT and inverse transform. That is, the first and the second inverse transform units 704 and 712 transform and invert the remaining pixels of the sub-block in the second and third inverse transform units 724 and 732, In order to allow only the residual signal Xd _i to be independently converted and inverse transformed, the modified basic vector shown in Figs. 10 and 11 is used.

For example, when one pixel of four pixels P _{0 to} P ₃ is transformed with the basic vector of the modified DCT transform shown in FIG. 10, a basic vector corresponding to the transformed first frequency coefficient Coeff ₀ 0, 0, 0, 2) reflects only the component of the arbitrary pixel residual signal Xd _i , the first frequency coefficient can be calculated as shown in Equation (4).

Therefore, when converting, using the basic vector of the modified DCT transform of FIG. 10, the first frequency coefficient has only the component of pixel P ₃ . Therefore, even if the values of the remaining pixels P _{0 to} P ₂ change, the values of the first frequency coefficients do not change, so that arbitrary pixels can be independently converted from the remaining pixels P _{0 to} P ₂ .

The first transform unit 708 transforms the arbitrary pixel residual signal Xd _i , which is a residual signal of an arbitrary pixel extracted from each sub-block in the current macroblock, independently from the remaining pixels of each sub-block, Two-dimensionally transformed into a basic vector (0, 0, 0, 2) corresponding to the frequency coefficient (Coeff ₀ ). Dimensional conversion is the same as multiplying the value of the arbitrary-pixel residual signal (Xd _i ) by 2, the two-dimensional conversion is equivalent to multiplying the value of the arbitrary-pixel residual signal (Xd _i ) by four. That is, the result of performing the two-dimensional conversion of each sub-block of the current macroblock to the basic vector of the modified DCT transform shown in FIG. 10, and then collecting only the frequency coefficient at the upper left of each sub-block. Accordingly, the first conversion unit 708 multiplies the value of the 16 (4x4) random position pixel residual signal Xd _i extracted from each sub-block in the current macroblock by 4, and performs two-dimensional conversion in the vertical direction and the horizontal direction Then, the transformed frequency coefficients are subjected to Hadamard transform using Equation (2), such as the frequency coefficients of the DC component of each sub-block of the intra 16x16 block of H.264.

Similarly, the frequency coefficient corresponding to the arbitrary pixel residual signal can be inversely transformed independently of the other frequency coefficients (Coeff _{1 to 3} ). The corresponding frequency coefficient at the time of inverse transform can be calculated as shown in Equation (5).

The first adder 716 adds the predictive arbitrary pixel block and the inverse-transformed arbitrary pixel residual block to generate a reconstructed arbitrary pixel block.

The second predictor 718 generates the predictive macroblock by predicting the remaining pixels of each sub-block using the reconstructed arbitrary pixels of the neighboring pixels of the current macroblock and the reconstructed arbitrary pixel block. That is, the second predictor 718 predicts each sub-block in the current macroblock, and predicts the remaining pixels excluding the arbitrary pixel in each sub-block.

FIGS. 12 to 15 illustrate exemplary prediction of the remaining pixels of each sub-block. As shown in FIGS. 12 to 15, the second predicting unit 718 predicts a pixel V _i ( _{i, j} ) adjacent to the current macroblock among the restored arbitrary pixel of the arbitrary pixel block which has been already coded and decoded and restored, V ₀ to V ₁₅ ) and H _i (H ₀ to H ₁₅ ) to the remaining pixels of each sub-block. In this case, the prediction direction may be the same as the prediction direction of the first prediction unit 704.

12, when the prediction direction is vertical, the remaining pixels located at the coordinates (0,0) to (0,15) in the current macroblock are H ₀ And the remaining pixels located at the coordinates (1,0) to (1,15) are predicted by the values of H ₁ And the remaining pixels located at coordinates (2,0) to (2,15) are predicted by the value of H ₂ Is predicted as the value of the pixel. However, in the case of the restored random pixel _{(X 0 ', X 4'} , X 8 ', X 12') of the pixel located in the region of the coordinates (3,0) ~ (3,15) passing through the coordinates (3, 0) to (3, 2), H ₃ The pixel located at the coordinates (3, 4) to (3, 6) is predicted as the reconstructed arbitrary pixel X ₀ ', and the pixel located at the coordinates (3, 8) The reconstructed arbitrary pixel X ₄ 'is predicted, and the pixel located at the coordinates (3, 12) to (3, 14) is predicted as the reconstructed arbitrary pixel X ₈ '. The remaining columns of the current macroblock are also predicted similar to the above.

13, in the case where the prediction direction is the horizontal direction, the coordinates (0,0) to (15,0), (0,1) to (15,1) in the current macroblock, , And (0,2) to (15,2) are V ₀ , V ₁ , V ₂ But the predicted value of the pixel, the restored random pixel _{(X 0 ', X 1'} , X 2 ', X 3') the passing interval of the coordinates (0,3) ~ For the pixel located in the (15,3) is , Pixels located at coordinates (0, 3) to (2, 3) are V ₃ Is predicted as the pixel value of the pixel located at the coordinates (4,3) ~ (6,3) is 'is predicted, the pixel located at the coordinates (8,3) - (10,3) is X _1' X ₀ to And the pixel located at the coordinates (12, 3) to (14, 3) is predicted as X ₂ '.

14, when the prediction direction is the DC average value prediction, the adjacent pixels V _i (V ₀ to V ₁₅ ) of the macroblock on the left side of the current macroblock and the adjacent pixels H _i (H ₀ to H ₁₅ ) And each reconstructed arbitrary pixel (X _i ') of each sub-block are given weight values, and an average value is calculated to perform prediction

Referring to FIG. 15, when the prediction direction is a plane direction, pixels located at coordinates (3, 0) to (3, 2) are predicted by assigning arbitrary weights to H ₃ pixel values and X ₀ ' , Pixels located at the coordinates (0, 3) to (2, 3) are V ₃ Pixel value and the prediction was given an arbitrary weight depending on the distance X ₀ 'value, a pixel with the coordinates (0,3) ~ located on the predicted coordinate (3,0) - (3,2) (2,3) (0,0) to (2,2) by assigning arbitrary weights to the restored arbitrary pixel X ₀ ', neighboring pixels V ₀ to V ₃ and H ₀ to H ₃ according to the distance, And predicts the located pixel. Other subblocks are similarly predicted.

The second subtractor 720 subtracts the predictive macroblock from the current macroblock of the original image to generate a residual macroblock. That is, the second subtractor 720 calculates residual signals of pixels other than the arbitrary-positioned pixel X _i in each sub-block in the current macroblock, and generates a residual macroblock having the residual signal.

The random residual pixel combiner 722 combines each inverse transformed residual pixel of the inverse transformed arbitrary pixel residual block into each position of the residual macroblock. That is, the arbitrary residual pixel combining unit 722 receives the inverse transformed random residual pixel (Xd _i ') in the inverse-transformed arbitrary pixel residual block by the inverse transform unit 714 from the second subtractor 720 And combines the generated residual macroblock with the position of the corresponding pixel of each sub-block. Therefore, although the residual macroblock generated by the second subtractor 720 includes only the residual signal of the remaining pixels, the residual residual macroblock reconstructed by the random residual pixel combiner 722 is combined, Is completed. As described above, the arbitrary pixel residual signal (Xd _i ') reconstructed at the position of the arbitrary residual pixel of each residual sub-block of the block of residual macroblock can be combined to reduce the quantization error that may occur according to the conversion and quantization. That is, since the quantized error is already reflected in the reconstructed arbitrary residual pixel, the reconstructed arbitrary pixel residual signal Xd _i 'can be encoded and decoded to the same value irrespective of the quantization error of the image encoding apparatus and the image decoding apparatus Do. Therefore, since only the quantization error of the remaining pixels is added in the encoding and decoding processes, the quantization error can be reduced as compared with the case of using the arbitrary pixel residual signal Xd _{i that} does not reflect the quantization error.

The second transform unit 724 transforms the residual macroblock to which the inverse transformed residual pixels are combined. That is, the second transform unit 724 transforms each residue signal of the residual macroblock combined with the inverse transformed residual pixels by the random residual pixel combining unit 722 into the frequency domain, and outputs the residual macroblock having the frequency coefficient .

Also, the second transforming unit 724 transforms the residual macroblocks in units of 4x4 blocks using the DCT transform, but can use another basic vector according to the order of the one-dimensional transform direction of the DCT transform. Here, the order of the one-dimensional conversion direction may be determined according to the prediction direction of the first prediction unit 704. [

FIG. 16 exemplarily shows one-dimensional transform for two-dimensional DCT transform. Referring to FIG. 16, the second transform unit 724 performs a one-dimensional transform in the horizontal direction first, a one-dimensional transform in the vertical direction to perform a two-dimensional DCT transform, or a one- Dimensional DCT transform by performing one-dimensional transform in the horizontal direction. The first predictor 704 determines whether to perform the one-dimensional transform in the horizontal direction first or the one-dimensional transform in the vertical direction first. May be determined according to a prediction direction in which an arbitrary pixel block is predicted.

The second conversion unit 724 can apply the basic vector used for the one-dimensional conversion differently according to the determined order of the one-dimensional conversion direction. 17 and 18 illustrate a process of performing inverse transformation using a modified basic vector according to the order of the inverse transformation directions. In the case where the one-dimensional vertical conversion is performed after the one-dimensional horizontal conversion is performed, the second conversion unit 724 converts the original basic vector into the first to third rows of each 4x4 block as shown in FIG. And the fourth row of each 4x4 block performs a one-dimensional horizontal transformation using the modified basic vector, the first column of each 4x4 block uses the modified basic vector and the second column to the fourth column of each 4x4 block The column can perform a one-dimensional vertical transformation using the modified basic vector.

When the one-dimensional horizontal conversion is performed after the one-dimensional vertical conversion is performed by the second conversion unit 724, as shown in FIG. 18, the first to third columns of each 4x4 block are converted into the basic Vector is used and the fourth column of each 4x4 block performs a one-dimensional vertical transformation using the modified basic vector, the first row of each 4x4 block is modified using the modified basic vector vector and the second row of each 4x4 block The fourth row can perform one-dimensional horizontal transform using the modified basic vector.

As described above, by determining the order of the directions of the one-dimensional transform according to the prediction direction and converting the basic vector used in the transformation according to the determined order by selectively using the existing basic vector or the modified basic vector, In addition, when the residual signal for the remaining pixels of each sub-block is transformed, it can be transformed independently of each of the arbitrary-pixel residual signals. Here, the existing basic vector and the modified basic vector are as described above with reference to FIG.

The second quantization unit 726 quantizes the transformed residual macroblock. That is, the second quantization unit 726 quantizes the frequency coefficients of the residual macroblocks transformed by the second transform unit 724 to generate quantized residual macroblocks, which are residual macroblocks having quantization frequency coefficients.

The encoding unit 728 encodes the quantized residual macroblock to generate a bitstream. That is, the encoding unit 728 generates a quantized frequency coefficient sequence by scanning the quantized frequency coefficients of the quantized residual macroblock using various scanning methods such as zigzag scanning, and generates a variety of encoding techniques such as entropy encoding, And outputs the bit stream. At this time, the encoding unit 728 can further encode information on the prediction mode transmitted from the first predictor 704 or the second predictor 718 as well as the quantized residual macroblock.

The second inverse quantization unit 730 dequantizes the quantized residual macroblock. That is, the second inverse quantization unit 730 dequantizes the quantization frequency coefficients of the residual macroblock quantized by the second quantization unit 726 to generate an inverse-quantized residual macroblock having the frequency coefficients.

The second inverse transform unit 732 inversely transforms the dequantized residual macroblock. That is, the second inverse transformer 732 inversely transforms the frequency coefficients of the residual macroblock inversely quantized by the second inverse quantization transformer 730 into a spatial domain, restores the arbitrary pixel residual signal, .

The second inverse transform unit 732 inversely transforms the dequantized residual macroblock by inverse transforming the frequency coefficients of the residual macroblocks in units of 4x4 blocks using the inverse DCT transform, The inverse transformation can be performed using another basic vector according to the order of the conversion directions. The second inverse transformation unit 732 can determine the order of the one-dimensional inverse transformation direction in the reverse order of the order of the one-dimensional transformation direction of the second transformation unit 724.

That is, when the one-dimensional vertical inverse transformation is performed after performing the one-dimensional horizontal inverse transformation, the second inverse transformation unit 732 transforms the first row of each 4x4 block into the modified basic Dimensional horizontal inverse transformation using the existing basic vectors, and the first to third columns of each 4x4 block use the existing basic vector and the second to fourth rows of the 4x4 block are subjected to the one- The fourth column of the 4x4 block can perform one-dimensional vertical inverse transform using the modified basic vector.

18, when the one-dimensional vertical inverse transform is performed and the one-dimensional horizontal inverse transform is performed, the second inverse transform unit 732 transforms the first row of each 4x4 block into a modified basic Dimensional vertical inverse transformation using the existing basic vector, and the first to third rows of each 4x4 block use the existing basic vector and the The fourth row of the 4x4 block can perform one-dimensional horizontal inverse transform using the modified basic vector.

The second adder 734 adds the inverse transformed residual macroblock to the predictive macroblock to generate a reconstructed current macroblock. That is, the second adder 734 adds each residual signal of the residual macroblock inversely transformed by the second inverse transform unit 732 and the respective pixels of the predictive macroblock predicted by the second predictor 718 Thereby generating a reconstructed current macroblock, which is a macroblock having reconstructed pixels.

The arbitrary reconstruction pixel combiner 736 combines each reconstructed arbitrary pixel of the reconstructed arbitrary pixel block with each position of the reconstructed current macroblock. That is, the arbitrary reconstructed pixel combiner 736 reconstructs each reconstructed arbitrary pixel of the arbitrary pixel block reconstructed by the first adder 716 to each sub-block of the current macroblock reconstructed by the second adder 734, To generate a current macroblock that is finally restored. The finally restored current macroblock is stored in a memory (not shown) and is used when the first predictor 704 and the second predictor 718 encode the next macroblock.

FIG. 19 is a flowchart illustrating an image encoding method according to an embodiment of the present invention.

When the image to be coded is input, the image coding device 700 divides the image into macroblock units, determines an optimal block mode among the various coding modes, and predicts and encodes macroblocks to be coded according to the determined block mode.

At this time, when the intra mode is determined as the encoding mode and intra prediction is performed, the image encoding apparatus 700 extracts arbitrary pixels from each sub-block of the current macro block to generate an arbitrary pixel block (S1910) And generates an arbitrary pixel residual block by subtracting the arbitrary pixel block from the predictive arbitrary pixel block (S1920).

The image encoding apparatus 700 transforms an arbitrary pixel residual block, quantizes the transformed arbitrary pixel residual block, inversely quantizes the quantized arbitrary pixel residual block, and inversely transforms the inverse quantized arbitrary pixel residual block (S1930) The arbitrary pixel block and the arbitrary pixel residual block inversely transformed are added to generate a reconstructed arbitrary pixel block (S1940). The image encoding apparatus 700 generates a predictive macroblock by predicting the remaining pixels of each sub-block using the reconstructed arbitrary pixels of the neighboring pixels of the current macroblock and the reconstructed arbitrary pixel block, The residual macroblock is generated by subtracting the block (S1950).

In addition, the image encoding apparatus 700 combines each inverse transformed residual pixel of the inverse-transformed arbitrary pixel residual block to each position of the residual macroblock (S1960) The quantized residual macroblock is transformed, and the quantized residual macroblock is encoded to generate a bitstream (S1970).

Then, as shown in FIG. 20, the image encoding apparatus 700 can restore the current macroblock. That is, the image encoding apparatus 700 dequantizes the residual macroblock quantized in step S1960 and inversely transforms the dequantized residual macroblock (S2010), adds the inverse transformed residual macroblock to the predictive macroblock, The current macroblock is generated (S2020), and each reconstructed arbitrary pixel of the reconstructed arbitrary pixel block is combined with each reconstructed current macroblock to generate a reconstructed current macroblock (S2030).

As described above, the image encoded by the image encoding apparatus 700 can be transmitted in real time or in non-real time through a wired / wireless communication network such as the Internet, a local area wireless communication network, a wireless LAN network, a WiBro network, And transmitted to a video decoding apparatus to be described later through a communication interface such as a serial bus (USB: Universal Serial Bus), decoded by the video decoding apparatus, and restored and reproduced as an image.

FIG. 21 is a block diagram schematically illustrating an electronic configuration of an image decoding apparatus according to an embodiment of the present invention. Referring to FIG.

The image decoding apparatus 2100 according to an embodiment of the present invention is an apparatus for decoding an image and includes a decoding unit 2102, an inverse quantization unit 2104, a coefficient extraction unit 2106, a first inverse transformation unit 2108, A first predictor 2110, a first adder 2112, a second inverse transformer 2114, a second predictor 2116, a second adder 2118, and a pixel combiner 2120 .

The decoding unit 2102 decodes the bitstream to extract residual macroblocks and prediction modes. That is, the decoding unit 2102 decodes the bitstream to extract the quantized frequency coefficient sequence and the prediction mode, and inversely scans the quantized frequency coefficient sequence using various inverse scanning methods such as inverse zigzag scanning to generate residual macroblocks.

The inverse quantization unit 2104 dequantizes the residual macroblock, and the coefficient extraction unit 2106 extracts the frequency coefficients corresponding to the positions of arbitrary pixels of each sub-block in the dequantized residual macroblock. Here, the frequency coefficient corresponding to the position of an arbitrary pixel may be the frequency coefficient of the DC component.

The first inverse transform unit 2108 generates an inverse transformed arbitrary pixel residual block by inversely transforming the frequency coefficients extracted by the coefficient extracting unit 2106. The method of inverse transforming the extracted frequency coefficients by the first inverse transform unit 2108 is the same as or similar to the inverse transform performed by the first inverse transform unit 714 described above with reference to FIG. 7, and thus a detailed description thereof will be omitted.

The first predictor 2110 predicts an arbitrary pixel block according to a prediction direction according to the prediction mode extracted by the decoding unit 2102 to generate a predictive arbitrary pixel block. The method of predicting an arbitrary pixel block according to the prediction direction by the first predicting unit 2110 is the same as or similar to the method predicted by the first predicting unit 704 through FIG. 7, and thus a detailed description thereof will be omitted. The first adder 2112 adds the predictive arbitrary pixel block and the inverse-transformed arbitrary pixel residual block to generate a reconstructed arbitrary pixel block.

The second inverse transform unit 2114 inversely transforms the dequantized residual macroblock to generate an inverse transformed residual macroblock. The method of inverse transforming the frequency coefficients of the inverse quantized residual macroblock by the second inverse transform unit 2108 is the same as or similar to the inverse transform performed by the second inverse transform unit 724 described above with reference to FIG. Is omitted.

The second predictor 2116 predicts the current macroblock of the image according to the prediction direction according to the prediction mode using the neighboring pixels of the current macroblock of the image and each reconstructed arbitrary pixel of the reconstructed arbitrary pixel block, Block. The second predictor 2116 predicts a current macroblock of an image according to a prediction direction according to a prediction mode using adjacent pixels of the current macroblock of the image and reconstructed arbitrary pixels of the reconstructed arbitrary pixel block, 7, the method is the same as or similar to the method predicted by the second predicting unit 718, and thus a detailed description thereof will be omitted.

The second adder 2118 generates the restored current macroblock using the prediction current macroblock and the inverse transformed residual macroblock. The arbitrary-pixel combining unit 2120 combines the reconstructed arbitrary pixel into each position of the reconstructed current macroblock. In this manner, the combined current macroblock becomes the last reconstructed current macroblock, and the reconstructed current macroblock is concatenated on a picture-by-picture basis and output as a reconstructed image. In addition, the finally restored current macroblock is utilized when the next macroblock is decoded by the first predictor 2110 and the second predictor 2116.

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

The video decoding apparatus 2100 receives and stores a bitstream of an image through a wired / wireless communication network or a cable, and decodes and restores the video in order to reproduce the video according to a user's selection or an algorithm of another program being executed.

To this end, the image decoding apparatus 2100 decodes the bitstream to extract the residual macroblock and the prediction mode (S2210), and dequantizes the extracted residual macroblock (S2220).

The image decoding apparatus 2100 extracts a frequency coefficient corresponding to a position of an arbitrary pixel of each sub-block in the dequantized residual macroblock (S2230), and generates a reconstructed arbitrary pixel residual block by inversely transforming the extracted frequency coefficient (S2240) and generates a predictive arbitrary pixel block by predicting the arbitrary pixel block according to the predictive direction according to the predictive mode (S2250). In operation S2250, the image decoding apparatus 2100 generates an arbitrary pixel block by adding the predictive arbitrary pixel block generated in operation S2250 and the reconstructed arbitrary pixel residual block in operation S2240.

The image decoding apparatus 2100 generates an inverse transformed residual macroblock by inversely transforming the dequantized residual macroblock (S2260), and reconstructs the reconstructed arbitrary pixel of the neighboring pixel of the current macroblock of the image and the reconstructed arbitrary pixel block The current macroblock of the image is predicted according to the prediction direction according to the prediction mode to generate a predictive current macroblock (S2270). The image decoding apparatus 2100 generates a reconstructed current macroblock using the predictive current macroblock and the inverse transformed residual macroblock (S2280), and combines the reconstructed arbitrary pixel with each position of the reconstructed current macroblock A restored image is generated and output (S2290).

Hereinafter, an image encoding apparatus according to another embodiment of the present invention will be described. The image encoding apparatus according to another embodiment of the present invention is the same as the image encoding apparatus 700 according to an embodiment of the present invention described above with reference to FIG. 7, but the functions of the image encoding apparatus are different.

That is, the image encoding apparatus 700 according to an embodiment of the present invention encodes and decodes each sub-block of the current macro block to be coded, and after coding all the sub-blocks, decodes all the sub-blocks Conversion and quantization are performed on a macroblock basis. However, in order to encode and decode all the sub-blocks of the current macroblock, the image encoding apparatus according to another embodiment of the present invention encodes and decodes one sub-block and then encodes and decodes the next sub- The subblock is coded and decoded. Therefore, according to another embodiment of the present invention, the transformation and the quantization are performed in units of subblocks.

Referring to FIG. 23, an image encoding apparatus 700 according to an embodiment of the present invention decodes each sub-block after completing encoding of each sub-block in a current macroblock in the order shown. That is, the image encoding apparatus 700 according to an embodiment of the present invention sequentially predicts the first subblock to the 16th subblock using adjacent pixels of the current macroblock and reconstructed arbitrary pixels of each subblock The quantized macroblocks are inversely quantized and inversely transformed, and each of the sub-blocks predicted by the transformed residual macroblocks is subtracted to obtain the quantized sub-blocks, Decodes the 16 sub-blocks.

However, in the image encoding apparatus according to another embodiment of the present invention, the second predictor 718 predicts the first sub-block using adjacent pixels and reconstructed arbitrary pixels, and the second subtractor 720 predicts Block, and the second transform unit 724 transforms the residual sub-block, and the second quantization unit 726 quantizes the transformed residual sub-block. The second inverse quantization unit 730 dequantizes the quantized residual sub-blocks, and the second inverse transform unit 732 adds the inverse quantized residual sub-blocks to the predicted sub-blocks, Restore the block. Subsequently, the second predictor 718 predicts the second sub-block. The adjacent pixel to be used for prediction is not the adjacent pixel of the current macroblock but the neighboring pixel on the left is the restored The neighboring pixel of the first sub-block is used, and the adjacent pixel at the upper side becomes the adjacent pixel of the current macroblock, and the second sub-block is predicted using the restored random pixels in the adjacent pixels and the second sub-block. The predicted 2 subblocks are subtracted from the 2 subblocks, and the residual subblocks are generated, transformed, quantized, inverse-quantized, and inverse transformed to restore the 2 subblocks. This process is performed in the order shown and each sub-block restored is used to encode each sub-block.

According to another aspect of the present invention, there is provided an image encoding apparatus for generating an arbitrary pixel block by extracting an arbitrary pixel in each sub-block of a current macroblock of an image, generating an arbitrary pixel block by predicting the arbitrary pixel block, An arbitrary pixel residual block is generated by subtracting the arbitrary pixel block and the predictive arbitrary pixel block, and then the arbitrary pixel residual block is transformed, the quantized arbitrary pixel residual block is quantized, the quantized arbitrary pixel residual block is dequantized, An arbitrary pixel residual block is inversely transformed and an arbitrary pixel block reconstructed by adding the predictive arbitrary pixel block and the inverse transformed arbitrary pixel residual block is generated. Then, the remaining pixels of each sub-block are predicted using adjacent pixels of each sub-block and each reconstructed arbitrary pixel of the reconstructed arbitrary pixel block to generate each predicted sub-block, and subtracting each predicted sub-block from the sub- Each residual sub-block is generated, each inverse-transformed residual pixel of the inverse-transformed arbitrary-pixel residual block is combined with each position of each residual sub-block, and each residual sub-block in which each inverse- Transforms each transformed residual sub-block, quantizes each quantized residual sub-block, and generates a bit stream by encoding each residual residual sub-block. At this time, one sub-block is coded and decoded in each sub-block, and the next sub-block is coded and decoded.

In addition, the image encoding apparatus according to another embodiment of the present invention dequantizes each quantized residual sub-block, inversely quantizes each residual sub-block inversely quantized, adds each inverse transformed residual sub-block to each predicted sub-block After reconstructed sub-blocks are generated, each reconstructed arbitrary pixel of the restored arbitrary pixel block is combined with each restored position of each sub-block. At this time, in predicting the remaining pixels of each sub-block, each restored sub-block is stored, and the next sub-block is predicted using each stored sub-block.

In addition, the image encoding apparatus according to another embodiment of the present invention does not use the prediction direction of the current macroblock or the prediction direction of the first prediction unit 704, but determines the prediction direction for each sub-block, The remaining pixels of the sub-block can be predicted. For this, the second predictor 718 of the image encoding apparatus according to another embodiment of the present invention estimates each reconstructed neighboring pixel in the neighboring pixels having the smallest value difference from each reconstructed arbitrary pixel among the neighboring pixels of each sub- The direction toward the arbitrary pixel can be determined as the prediction direction. In this case, the second predictor 718 according to another embodiment of the present invention may generate information on the prediction mode according to the prediction direction for each sub-block, and may transmit or not transmit information to the encoder 728 . When the information on the prediction mode for each sub-block is transmitted to the encoding unit 728, the encoding unit 728 encodes the information on the prediction mode for each sub-block, and outputs the prediction mode for each sub- And the prediction direction can be synchronized using the information on the prediction mode when the image decoding apparatus predicts each sub-block. If the information on the prediction mode for each sub-block is not coded, the image decoding apparatus determines the prediction direction of each sub-block and predicts it in the determined prediction direction as in the image coding apparatus.

Referring to FIG. 24, the second predictor 718 according to another embodiment of the present invention may include a prediction unit 704 for estimating a difference between the left (V _i ) and the upper (H _i ) First, the reconstructed pixels are compared with the reconstructed arbitrary pixel (X _i ') to determine a new optimum prediction direction of each sub-block, and then each sub-block is predicted according to the determined new optimum prediction direction. The new optimal prediction direction of each sub-block is determined by the direction of the edge (edge) by using the arbitrary pixel (X _i ) and the neighboring pixels (V ₀ to V ₃ , H ₀ to H ₇ ) Is determined by prediction.

For example, if the values of X _i and H ₃ are the closest, the optimal prediction direction of the sub-block is determined to be the vertical direction, and if the values of X _i and V ₃ are closest to each other, And is determined in the horizontal direction. The DC prediction mode is determined when V ₀ to V _{3 ,} H ₀ to H ₇ and X _i values are similar. The left diagonal is determined to be similar to V ₀ and H ₀ , V ₁ and H ₁ , V ₂ and H ₂ , V ₃ and H _3, and H ₇ and X _i values. The diagonal right direction is determined when the values of N and X _i are similar. Vertical right direction is determined when H ₁ and X _i values are similar. Horizontal downward direction is determined when V ₂ and X _i values are similar. Vertical leftward direction is determined when H ₅ and X _i are similar, H ₀ and V ₁ , H ₁ and V _2, and H ₂ and V ₃ are similar. Horizontal upward direction is determined when H ₁ and V ₀ , H ₂ and V ₁ , H ₃ and V _2, and H ₄ and V ₃ are similar. As described above, the second predicting unit 718 according to another embodiment of the present invention can increase the accuracy of prediction by determining and predicting the optimum prediction direction for each sub-block. That is, since the rest of the pixels are predicted using the reconstructed arbitrary pixel for each sub-block, the optimal prediction direction can be determined for each sub-block unlike the conventional prediction method depending on the relationship between the reconstructed arbitrary pixel and the adjacent pixels. It is possible to improve compression efficiency through encoding and decoding through switching of the same prediction direction.

Here, the order in which the image encoding apparatus according to another embodiment of the present invention encodes and decodes each sub-block may be a sequence of encoding and decoding an intra 4x4 macroblock of H.264 / AVC or a raster scan direction . The procedure shown in Fig. 23 shows a procedure for coding and decoding intra 4x4 macroblocks in the H.264 / AVC standard, and Fig. 25 shows the coding / decoding procedure for each sub-block in the raster scanning direction. As shown in the figure, the image encoding apparatus according to another embodiment of the present invention restores and restores one sub-block by predicting, transforming, quantizing, inverse-quantizing, Block, and performs decoding, prediction, conversion, quantization, inverse-quantization, and inverse transform using the sub-blocks.

FIG. 26 exemplarily shows how the remaining pixels of each sub-block are predicted according to a newly determined optimal prediction direction using the reconstructed arbitrary pixel X and adjacent pixels H, V, and N. FIG. In another embodiment of the present invention, the remaining pixels of each sub-block are predicted in the manner as shown.

In addition, the image encoding apparatus according to another embodiment of the present invention includes all the other features of the image encoding apparatus 700 according to an embodiment of the present invention. That is, the first transform unit 708 of the image encoding apparatus according to another embodiment of the present invention can independently convert and inverse transform only the arbitrary pixel residual signals of the arbitrary pixel residual block, Only the arbitrary pixel residual signal can be independently transformed by performing discrete cosine transform using the corrected pixel basic signal, and Hadamard transform can be further performed after discrete cosine transform.

In addition, the first inverse transform unit 714 of the image encoding apparatus according to another embodiment of the present invention performs inverse discrete cosine transform using the corrected basic vector, Only the pixel residual signal can be independently inversely transformed, and the inverse discrete cosine transform can be performed after inverse Hadamard transform of the transformed frequency coefficient of each arbitrary pixel residual signal.

The second transform unit 724 of the image encoding apparatus according to another embodiment of the present invention transforms the residual macroblocks in 4x4 block units using the discrete cosine transform, You can then use another basic vector. At this time, the order of the one-dimensional conversion direction may be determined according to the prediction direction of the first predictor. When the one-dimensional vertical conversion is performed after performing the one-dimensional horizontal conversion, the second conversion unit 724 uses the existing basic vectors in the first to third rows of each 4x4 block, The fourth row performs a one-dimensional horizontal transform using the modified basic vector, the first column of each 4x4 block uses the modified basic vector, and the second to fourth columns of each 4x4 block are modified basic vectors Dimensional vertical transformation can be performed using the conventional basic vector. When performing the one-dimensional vertical transformation after performing the one-dimensional vertical transformation, the first to third columns of each 4x4 block use the existing basic vector, The fourth column of the 4x4 block performs a one-dimensional vertical transformation using the modified basic vector, the first row of each 4x4 block uses the modified basic vector, and the second through fourth rows of each 4x4 block are modified Basic vector < RTI ID = 0.0 > You can perform a circle horizontal conversion.

In addition, the second inverse transform unit 732 of the image encoding apparatus according to another embodiment of the present invention inverts the frequency coefficients in units of 4x4 blocks using inverse discrete cosine transform, and performs a one-dimensional inverse transform of the inverse discrete cosine transform The order of the one-dimensional inverse transformation directions can be determined in the reverse order of the order of the one-dimensional transformation directions of the second transformation unit 724. When the one-dimensional vertical inverse transform is performed after performing the one-dimensional horizontal inverse transform, the second inverse transform unit 732 uses the modified basic vector in the first row of each 4x4 block, The second to fourth rows perform a one-dimensional horizontal inverse transformation using the existing basic vector, and the first to third columns of each 4x4 block use the existing basic vector and the fourth column of each 4x4 block is modified Dimensional vertical inverse transformation can be performed using the basic vector, and in the case of performing the one-dimensional vertical inverse transformation after performing the one-dimensional vertical inverse transformation, the first column of each 4x4 block is transformed into the modified basic vector Dimensional vertical inverse transformation using the existing basic vector, and the first to third rows of each 4x4 block are generated by using the existing basic vector and each 4x4 block, The fourth row of < RTI ID = 0.0 > Using the emitter may perform a one-dimensional horizontal inverse transform. In addition, the second inverse transform unit 732 can inversely transform the frequency coefficients in units of 4x4 blocks using the inverse discrete cosine transform, and can use the existing basic vectors. The second inverse transform unit 732 transforms a part of the frequency coefficients, The remaining frequency coefficients among the frequency coefficients can be predicted by using a part of pixels.

In addition, the first predictor 704 according to another embodiment of the present invention may predict an arbitrary pixel block using pixels that are subjected to low-pass filtering and downsampled on the adjacent pixels of the current macroblock, It may be predicted in the same prediction direction as the prediction direction according to the prediction mode.

Meanwhile, in the image encoding apparatus according to the embodiment of the present invention and the image encoding apparatus according to the other embodiment of the present invention, the second inverse transform unit 732 transforms a part of the frequency coefficients, It is possible to predict the remaining frequency coefficients of the frequency coefficients using a part of the pixels and to invert the arbitrary pixel independently of the other pixels while performing inverse DCT transformation using the existing basic vector without using the modified basic vector have.

In such a two-dimensional inverse transform, the second inverse transform unit 732 performs two-dimensional transform using the DCT inverse transformed basic vector shown in FIG. 5, that is, the existing basic vector, instead of the modified basic vector. After the inverse transformation, the quantized frequency coefficient is substituted for the frequency coefficient of the DC component and transmitted.

FIGS. 27 and 28 illustrate a process of predicting remaining frequency coefficients using a part of frequency coefficients according to an embodiment of the present invention and another embodiment.

When the first transform unit 724 first transforms one-dimensional transforms in the horizontal direction and then vertically transforms one-dimensional transforms, the second inverse transform unit 732 can calculate DC _P using Equation (6). Equation (6) is a mathematical expression for performing one-dimensional inverse DCT using the existing basic vector for inverse transformation shown in FIG.

DC ₀₀ shown in equation (1) is a frequency coefficient transformed by the first transform unit 708, and the remaining AC ₀₁ through AC ₃₃ frequency coefficients are frequency coefficients after being transformed by the second transform unit 724, If the reconstructed arbitrary pixel X _i 'is the same as (2), AC 1, AC 2, and AC 3 are obtained by performing one-dimensional inverse DCT on only the first through third columns in the vertical direction as shown in (3) ) calculates a DC _P using AC①, AC②, AC③, the restored random pixels (X _i '), such as.

Summarizing Equation (6) with respect to Coef (1), Equation (7) can be derived.

Here, Coef②, Coef③, and the Coef④ 27 and AC①, AC②, AC③ in Fig. 28, respectively, and the DC Coef① _P.

Then, calculate DC ₀₀ p with AC ₁₀ , AC ₂₀ , AC ₃₀ and DCp as in (5). The method of calculating DC ₀₀ p is the same as the method of calculating DC _P described above. Using the calculated DC ₀₀ p and AC ₁₀ , AC ₂₀ , and AC ₃₀ , the remaining one column is subjected to one-dimensional inverse transformation in the vertical direction as shown in (6). The one-dimensional inverse transform of the other horizontal direction is also inversely transformed as shown in (7) using the DCT inverse basic vector of FIG. 5, that is, the existing basic vector, and the restored pixel of the corresponding sub-block as shown in (8) is output. By using such a method, the value of the frequency coefficients of the AC component can be reduced as compared with the case of using the modified basic vector as shown in FIG. Further, the arbitrary pixel can be inversely transformed independently of the frequency coefficients of the remaining pixels or the remaining pixels of the sub-block using the existing basic vector instead of the modified basic vector.

While the present invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not limited to the disclosed embodiments. That is, within the scope of the present invention, all of the components may be selectively coupled to one or more of them. In addition, although all of the components may be implemented as one independent hardware, some or all of the components may be selectively combined to perform a part or all of the functions in one or a plurality of hardware. As shown in FIG. The codes and code segments constituting the computer program may be easily deduced by those skilled in the art. Such a computer program can be stored in a computer-readable storage medium, readable and executed by a computer, thereby realizing an embodiment of the present invention. As the storage medium of the computer program, a magnetic recording medium, an optical recording medium, a carrier wave medium, or the like may be included.

It is also to be understood that the terms such as " comprises, "" comprising," or "having ", as used herein, mean that a component can be implanted unless specifically stated to the contrary. But should be construed as including other elements. All terms, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined. Commonly used terms, such as predefined terms, should be interpreted to be consistent with the contextual meanings of the related art, and are not to be construed as ideal or overly formal, unless expressly defined to the contrary.

The foregoing description is merely illustrative of the technical idea of the present invention, and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are intended to illustrate rather than limit the scope of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention.

As described above, the present invention is applied to an apparatus and a method for encoding or decoding an image, and it is an object of the present invention to improve the accuracy of prediction when intraprediction of pixels of a block to be currently encoded or decoded, It is a very useful invention.

Claims (7)

A method for intraprediction of a current block in units of subblocks,
A prediction mode determining step of determining an intra prediction mode of the current block; And
And restoring a plurality of subblocks constituting the current block sequentially according to an intra prediction mode of the determined current block,
Wherein, in the sub-block reconstructing step,
Generating predicted subblocks using pixels of neighboring subblocks predicted and restored according to an intra prediction mode of the determined current block in the current block;
Transforming the frequency-transformed residual signal of the restored sub-block into a spatial domain to generate residual sub-blocks; And
And summing the predicted sub-block and the residual sub-block to reconstruct the sub-block,
Wherein the generating of the residual sub-block comprises: transforming the frequency-
Performing one-dimensional transform on each column of the frequency-converted residual signal; And
Performing one-dimensional transform on each row of the transformed residual signal for each column;
The intra prediction method comprising: The method according to claim 1,
And the neighboring sub-blocks are located at the upper or left side of the restoration target sub-block. delete delete The method according to claim 1,
Wherein a plurality of subblocks constituting the current block coincide with a conversion unit in which a residual signal of the current block is transformed, The method according to claim 1,
Wherein a plurality of intra-prediction modes associated with the intra-prediction direction are determined by calculating a formula for predicting a sub-block by selectively combining pixels of neighboring sub-blocks.

delete

Images