KR100728032B1 - Method for intra prediction based on warping - Google Patents

Abstract

An intra-prediction method based on warping is provided to improve accuracy of intra-prediction of a region having a specific pattern and encoding efficiency by using an intra-prediction method employing a warping technique for intra-prediction for encoding and decoding images. A first intra-prediction is performed for blocks each having a predetermined size. A block group composed of a plurality of blocks each having a predetermined size is formed(920). A second intra-prediction is carried out for the block group(940). Bit rates according to the first intra-prediction and the second intra-prediction are calculated. One of the first intra-prediction and the second intra-prediction is selected on the basis of the bit rates.

Description

Warping-based intra prediction method {Method for intra prediction based on warping}

1 is a block diagram showing a conventional encoding apparatus.

2 is a block diagram showing a conventional decoding apparatus.

3A is a diagram illustrating types of prediction modes.

3B is a diagram illustrating prediction directions of prediction modes.

4 is a diagram illustrating pixels and a direction used when performing intra prediction on nine modes.

5 is a diagram illustrating peripheral pixels required to perform a 4x4 prediction mode.

6 is a diagram illustrating an encoding apparatus to which an intra prediction method is applied according to the present invention.

7 is a view for explaining a search region in which the intra prediction method is performed according to the present invention.

8 is a view for explaining an intra prediction method according to the present invention.

FIG. 9 is a flowchart for describing an intra prediction method performed by the encoding apparatus of FIG. 6.

10 is a block diagram illustrating an encoding apparatus according to another embodiment of the present invention.

FIG. 11 is a flowchart for describing an intra prediction method performed by the encoding apparatus of FIG. 10.

The present invention relates to video encoding and decoding, and more particularly, to a warping-based intra prediction method.

The popular MPEG video compression standard, H.264 or MPEG AVC, uses various compression techniques. Unlike conventional coding standards, various techniques for increasing compression efficiency, such as motion compensation using multiple reference frames, loop filtering, motion compensation based on variable size blocks, and entropy coding such as context adaptive arithmetic coding (CABAC) Adopt them. On the other hand, in order to improve the compression efficiency of the intra block, a prediction method in the spatial domain is adopted, which is called intra prediction, which significantly increases the overall coding efficiency.

1 is a block diagram showing a conventional encoding apparatus. Examples of the video encoding system shown in FIG. 1 include MPEG 2, MPEG 4, and an H.264 encoder.

First, the input image data is divided in units of 16 × 16 macroblocks.

The encoder controller 110 performs a function of determining a quantization coefficient of each block so as to satisfy a target bit for each picture and a bit rate of the entire desired sequence as a bit rate controller.

The transform and quantization unit 120 transforms the input image data in order to remove spatial redundancy of the image data. Further, transform coefficient values obtained by transform encoding are quantized according to a predetermined quantization step to obtain N × M data, which is two-dimensional data composed of quantized transform coefficient values. An example of an image transformation to be used is a discrete cosine transform (DCT). Quantization is performed according to a predetermined quantization step.

The inverse transform and inverse quantization unit 130 inverse quantizes the quantized image data in the transform and quantization unit 120, and inverse image transforms, for example, inverse DCT the inverse quantized image data.

The de-blocking filter 140 performs filtering to remove a blocking phenomenon caused by quantization in the motion compensated image, and outputs the result to the frame memory unit 150.

The frame memory unit 150 stores the image data inversely quantized and inversely transformed by the inverse transform and inverse quantization unit 130 in units of frames.

In the case of the intra macroblock, the intra frame predictor 160 obtains a predictor for each block or macroblock in the spatial domain, subtracts the predictor from the intra macroblock, and sends the difference to the transform unit.

The motion prediction and compensator (ME / MC) 170 estimates the motion vector per macroblock (MV) and SAD by using the input image data of the current frame and the image data of the previous frame stored in the frame memory unit 150. . Further, a motion compensated prediction region P, for example, a 16 × 16 region selected by motion estimation, is generated based on the estimated motion vector.

The entropy encoder 180 receives information about the quantized transform coefficients output from the transform and quantization unit 120 and the motion vector output from the motion prediction and compensation unit 170, and is provided by the encoder controller 110. Coding type information, quantization step information, and other information necessary for decoding are input and entropy-encoded to finally output a bitstream obtained.

That is, according to the encoding apparatus of FIG. 1, the adder 190 generates a residual image by subtracting the motion compensated prediction region P generated by the motion predictor and compensator 170 from the input current macroblock. The generated residual image is orthogonally transformed and quantized such as DCT in the transform and quantization unit 120. The entropy encoder 180 entropy-encodes header information such as coefficients, motion information, and control data of a macroblock output from the transform and quantizer 120 to generate a compressed bitstream.

2 is a block diagram showing a conventional decoding apparatus.

The decoding apparatus shown in FIG. 2 includes an entropy decoding unit 210, an inverse quantization and inverse transform unit 220, a frame memory unit 230, and a motion compensator 240.

The entropy decoder 210 entropy decodes the encoded input stream to extract image data, a motion vector, and the like. Entropy-decoded image data is input to the inverse quantization and inverse transform unit 220, and motion vector information is input to the motion compensator 240.

For example, the inverse transform and inverse quantization is performed by the inverse transform and inverse quantization unit 220, and the image data is reconstructed by adding a predictor, for example, a motion compensated prediction region value in the motion compensator 240. Is output to a display unit (not shown).

In the encoder and the decoder, the intra prediction unit is a core module. In H.264 intra prediction, 4x4, 8x8, and 16x16 modes exist according to a block size applied to luminance Luma, and a separate mode exists for chroma.

For example, in the case of the 4x4 mode applied to the luminance image, there are nine 4x4 prediction modes according to the prediction direction as shown in FIG. 3. 3 (a) shows the modes of the prediction modes. 3 (b) shows the prediction directions of the prediction modes.

4 illustrates pixels and directions used when performing intra prediction on nine modes. The light colored 4x4 is the current block to which intra prediction is to be applied, and the dark colored pixels are reconstructed neighboring pixels to be used for prediction. For example, in the vertical mode, a residual block is formed by forming a 4x4 predictor in a pattern in which four pixel values on the current block are vertically repeated, and correspondingly obtaining a difference from pixel to pixel of the current block. block) and then encoded through the residual block transform and quantization, CABAC, and the like.

For example, when the upper neighboring block pixel value is {255,0,255,0}, the 4x4 prediction block is as follows. In the same way

{255,0,255,0,

0,255,0,255,

255,0,255,0,

0,255,0,255}

The same applies to the horizontal mode. The other modes also have some equation for generating the predictor 4x4. Meanwhile, in order to perform 4x4 prediction, the upper nine pixels and the four reconstructed pixels on the left side are needed, as shown in FIG. 5, which shows peripheral pixels required to perform the 4x4 prediction mode.

In this way, even if the encoder performs intra prediction by generating patterns for various directions, since the pixel values used to generate the predictor are pixels of the neighboring block rather than the pixels in the current block, the prediction is limited. There was also a problem of low efficiency.

SUMMARY OF THE INVENTION The present invention has been made in an effort to provide an improved intra prediction method that solves such problems in the conventional encoder and decoder.

In order to achieve the above object, the intra prediction method according to the present invention comprises the step of forming a block group consisting of a plurality of blocks of a predetermined size, and performing the intra prediction in units of the formed block group, the intra Prediction is performed while varying the direction, size, or direction and size of the block group.

Further, preferably, in the intra prediction method according to the present invention, the predetermined size block is a macro block, and the intra prediction is performed while rotating and zooming the macro block.

The technical problem may further include performing first intra prediction on a block basis of a predetermined size; Forming a block group consisting of a plurality of blocks of a predetermined size; Performing second intra prediction on the basis of the formed block group; Calculating bit amounts according to the first intra prediction and the second intra prediction; Based on the calculated bit amount, it is also achieved by an intra prediction method comprising selecting one of a first intra prediction and a second intra prediction.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

6 is a diagram illustrating an encoding apparatus to which an intra prediction method is applied according to the present invention. Hereinafter, the present invention will be described with reference to FIGS. 7 and 8. 7 is a diagram illustrating a search region according to the present invention, and FIG. 8 is a diagram for explaining an intra prediction method according to the present invention.

The encoding apparatus according to the present invention includes a superMB generator 620, an intra predictor 640, and an encoder 660.

The superMB generator 620 generates a superMB. superMB contains one or more macroblocks (MBs). Here, the size of the macroblock is 16x16. In this embodiment, MB of size M × N is defined as superMB. Optionally, it is also possible to apply the present embodiment to a block group consisting of blocks having other sizes than the macroblocks.

In FIG. 7, the size of the current superMB is 3x4, and the MB of the position indicated by the asterisk is the current MB in the coding order.

The intra prediction unit 640 performs intra prediction according to the present invention in units of superMBs. As shown in FIG. 7, the search area in the same frame of the current superMB is a casual area, that is, a reconstructed area from the current MB. This is a search area of the largest possible size of the frame shown in FIG. Optionally, it is also possible to define a search area. Once the search area is determined, as shown in FIG. 8, the reference superMB (or predictor) corresponding to the current superMB is searched. In this case, as shown in FIG. 8, the predictor is searched by applying a warping technique to the current superMB. The warping technique searches for matching predictors by rotating the current superMB or changing the size or shape of the search. To this end, a motion estimation compensation scheme based on a 6 parameter model and an 8 parameter model is used.

The 6-parameter model is also known as the affine model, where the coordinates before transformation (x, y) and the coordinates after transformation (x ', y') have the following relationship.

(x ', y') = (ax + by + c, dx + ey + f)

A, b, c, d, e, f Six parameters define the motion between two coordinates.

On the other hand, the 8-parameter model is also referred to as a perspective model, a case in which the pre-conversion left (x, y) and the post-conversion coordinate (x ', y') have a relationship as shown in Equation 1 below.

Thus, the motion between two coordinates is defined by eight parameters a b, c, d, e, f, g, h.

In addition to these motion estimation compensation methods, it is also possible to use another method using the motion estimation compensation method using the zooming technique to change the rotation or the magnitude.

The encoder 660 obtains a residual image of the predictor obtained by motion compensation and the current superMB. As shown in FIG. 1, encoding of the differential image is performed by performing image transformation, quantization, and variable length encoding.

In this embodiment, the difference image of superMB is performed in MB units.

FIG. 9 is a flowchart for describing an intra prediction method performed by the encoding apparatus of FIG. 6.

In step 920, a superMB including one or more macroblocks (MB) is generated.

In step 940, a reference superMB (or predictor) is searched for using the warping technique in units of superMBs. The warping technique searches for matching predictors by rotating the current superMB or changing the size or shape of the search.

In step 960, a differential image of a predictor obtained by motion compensation and a current superMB is calculated, and image transform, quantization, and variable length encoding are performed on the calculated differential image, as shown in FIG. 1. Is done.

10 is a block diagram illustrating an encoding apparatus according to another embodiment of the present invention.

The encoding apparatus according to the present invention includes a first intra predictor 1020, a superMB generator 1040, a second intra predictor 1060, and an encoder 1060.

In the present embodiment, since the first intra prediction unit 1020 performs intra prediction according to the intra prediction method used in the conventional encoding and decoding apparatus, detailed description is omitted for simplicity.

In addition, since the superMB generator 1040 and the second intra predictor 1060 perform the same functions as the superMB generator 620 and the intra predictor 640 of FIG. 6, a detailed description thereof will be omitted.

The encoder 1080 compares the data obtained by the first intra predictor 1020 with the data obtained by the second intra predictor 1060 to determine whether to use the data obtained by the second intra predictor 1060. Decide

For example, when it is determined that the accuracy of intra prediction performed by the second intra prediction unit 1060 is relatively low, encoding is performed in units of MB according to a conventional encoding method.

In addition, when it is determined that the accuracy of the intra prediction performed by the second intra prediction unit 1060 is high, the MBs in the superMB are encoded in a predetermined scan method, for example, raster scanning order, and remainder in the picture except these. MBs encode in a raster scanning order as in the conventional encoding scheme in terms of the entire picture. At this time, the motion parameter information is optionally transmitted along with the position information of the leftmost macroblock of the corresponding superMB.

The intra prediction method according to the present invention may be performed in the decoding apparatus in the same manner only by changing a predetermined order.

FIG. 11 is a flowchart for describing an intra prediction method performed by the encoding apparatus of FIG. 10.

In operation 1120, intra prediction according to an intra prediction method used in a conventional encoding and decoding apparatus is performed.

In operation 1140, a superMB including one or more macroblocks (MBs) is generated.

In step 1160, a reference superMB (or predictor) is searched using a warping technique in units of superMBs. In the warping technique, a searcher is searched for a matching predictor by rotating the current superMB, changing its size or shape, and calculating a differential image of the predictor obtained by motion compensation and the current superMB.

In operation 1180, the data obtained in operation 1160 and the data obtained in operation 1160 are compared to determine whether to use the data obtained in operation 1160.

For example, when it is determined that the accuracy of the intra prediction performed in step 1160 is relatively low, encoding is performed in MB units according to a conventional encoding method based on the intra prediction result performed in step 1120.

In addition, when it is determined that the accuracy of the intra prediction performed in step 1160 is high, the MBs in the superMB are encoded in a predetermined scan method, for example, raster scanning order, and the remaining MBs in the picture except these are in view of the entire picture. As in the conventional encoding method, encoding is performed in the raster scanning order.

The intra prediction method according to the present invention may be performed in the decoding apparatus in the same manner only by changing a predetermined order.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

As described above, in the intra prediction for image encoding and decoding, by using the intra prediction method according to the present invention employing the warping technique, it is possible to improve the accuracy of intra prediction in the region having a specific pattern and the coding efficiency accordingly. There is an effect that it is possible.

Claims (4)

In the intra prediction method, Forming a block group consisting of a plurality of blocks of a predetermined size; Performing intra prediction on the basis of the formed block group; The intra prediction method is performed by changing the direction, size, or direction and size of the block group. The method of claim 1, The predetermined size block is a macroblock, and the intra prediction is performed while rotating and zooming the macroblock. In the intra prediction method, Performing first intra prediction on a block basis of a predetermined size; Forming a block group consisting of a plurality of blocks of a predetermined size; Performing second intra prediction on the basis of the formed block group; Calculating bit amounts according to the first intra prediction and the second intra prediction; And selecting one of a first intra prediction and a second intra prediction based on the calculated bit amount. The method of claim 3, The predetermined size block is a macroblock, and the intra prediction is performed while rotating and zooming the macroblock.

Images