JP2008219360A - Predictive encoding device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a predictive encoding device capable of performing moving image compression processing at a high speed. <P>SOLUTION: The predictive encoding device is provided with: an original image converting part for generating an original image having a first frequency conversion pattern by performing the first frequency conversion of the original image; and a predicted image converting part for generating a predicted image having the first frequency conversion pattern by performing the first frequency conversion of the predicted image. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a predictive coding apparatus.

  In recent years, various predictive coding apparatuses have been developed for the purpose of performing predictive coding, which is one of the moving picture compression techniques, and for example, an H.264 / AVC encoder has been developed as a representative example. However, in such a predictive coding apparatus, when determining an optimal prediction mode, it is possible to improve the prediction performance by obtaining an absolute value error sum obtained by performing orthogonal transformation such as Hadamard transform, but performing orthogonal transformation. The prediction mode determination has a problem that a large amount of processing is required and it is difficult to perform high-speed processing.

The following is a list of literature names related to the predictive coding apparatus.
T. Wiegand, GJ Sullivan, G. Bjontegaard, and A. Luthra, “Overview of the H.264 / AVC Video Coding Standard”, IEEE Trans. On Circuits and Systems for Video Technology, vol. 13, no. 7, pp . 560-576, July 2003 IG Richardson, H.264 and MPEG-4 Video Compression, Wiley, 2003 Joint Video Team (JVT) of ITU-T VCEG and ISO / IEC MPEG, Draft ITU-T Recommendation and Final Draft International Standard of Joint Video Specification, ITU-T Rec.H.264 and ISO / IEC 14496-10 AVC, May 2003

  An object of the present invention is to provide a predictive coding apparatus capable of performing moving image compression processing at high speed.

A predictive coding apparatus according to an aspect of the present invention includes:
A prediction image creating unit that creates a prediction image from a reference image based on the prediction mode selected from a plurality of different prediction modes;
An original image conversion unit for generating the original image having the first frequency conversion pattern by performing a first frequency conversion on the original image;
A prediction image conversion unit that generates the prediction image having the first frequency conversion pattern by performing the first frequency conversion on the prediction image;
A residual image having the first frequency conversion pattern is generated by calculating a difference between the original image having the first frequency conversion pattern and the predicted image having the first frequency conversion pattern. A first residual image conversion unit;
Based on the residual image having the first frequency conversion pattern, a mode determination unit that determines the prediction mode in which the code amount is small from the plurality of prediction modes;
A second residual image conversion unit that generates the residual image having a second frequency conversion pattern by performing a second frequency conversion on the residual image of the original image and the predicted image;
An image quality adjustment unit that performs image quality adjustment on the residual image having the second frequency conversion pattern;
The residual image for generating the residual image in the real space is obtained by performing inverse transformation of the second frequency transformation on the residual image having the second frequency transformation pattern that has undergone the image quality adjustment. A difference image inverse transform unit;
A reference image creating unit that creates the reference image using the residual image and the predicted image and outputs the reference image to the predicted image creating unit;

  According to the predictive coding apparatus of the present invention, moving picture compression processing can be performed at high speed.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 shows the configuration of a predictive coding apparatus 10 according to an embodiment of the present invention. The predicted image creation unit 140 performs intra-screen predictive coding that performs prediction using surrounding pixels based on the reference image output from the reference image creation unit 150 and the prediction mode output from the mode selection unit 80. As a result, a predicted image is created and output to the predicted image Hadamard transform unit 90, the residual image creating unit 100, and the reference image creating unit 150.

のように表されると、2次元アダマール変換は、

と表される。 An original image Hadamard transform unit 50 as an original image transform unit performs, for example, two-dimensional Hadamard transform (first frequency transform) on the original image, and has the original image (first frequency transform pattern that has been subjected to the Hadamard transform). The original image) is output to the residual image Hadamard transform unit 60. For example, an arbitrary matrix X with 4x4 elements and Hadamard transformation matrix T

The two-dimensional Hadamard transform is expressed as

It is expressed.

  The predicted image Hadamard transform unit 90 as the predicted image transform unit performs, for example, 0, 1 or two-dimensional Hadamard transform (first frequency transform) on the predicted image, and this Hadamard transformed predicted image (first The predicted image having the frequency conversion pattern) is output to the residual image Hadamard transform unit 60. Note that the dimensions of the Hadamard transform are selectively used according to the nature of the predicted image, that is, the substantial number of dimensions of the predicted image. For example, if the predicted image is 0-dimensional as shown in FIG. 2 (a), 0-dimensional Hadamard transform, if it is one-dimensional as shown in FIG. 2 (b), 1-dimensional Hadamard transform, as shown in FIG. 2 (c). If it is two-dimensional, the two-dimensional Hadamard transform is used.

  Note that, as the first frequency transform, various other first orthogonal transforms may be used instead of the Hadamard transform.

  The residual image Hadamard transform unit 60 as the first residual image transform unit calculates the difference between the Hadamard-transformed original image and the Hadamard-transformed predicted image, thereby obtaining a Hadamard transform pattern ( (Residual image having the first frequency pattern) is derived and output to the SATD calculator 70.

  A SATD (Sum of Absolute Transformed Difference) calculation unit 70 calculates the sum of absolute values of each element of Hadamard transformed data. Since this total sum, that is, the SATD is small, the amount of code generated by predictive coding is expected to be small, so it is used as an index for determining the prediction mode.

  The mode selection unit 80 selects an optimal prediction mode with a small code amount from a plurality of different prediction modes based on SATD, and outputs this to the prediction image creation unit 140. The SATD calculation unit 70 and the mode selection unit 80 form a mode determination unit 40.

  The residual image creation unit 100 calculates a residual between the original image and the predicted image and outputs the obtained residual image to the residual image DCT unit 110.

のように表されると、2次元DCTは、

と表される。 The residual image DCT unit 110 performs, for example, DCT (Discrete Cosine Transform) (second frequency transformation) on the residual image, and the DCT residual image (residual image having the second frequency transformation pattern). Difference image) is output to the image quality adjustment unit 120. For example, an arbitrary matrix X having 4 × 4 elements and a DCT transformation matrix C are expressed by the following equations:

The two-dimensional DCT is expressed as

It is expressed.

  As the second frequency transform, various other second orthogonal transforms using a transform matrix different from the first orthogonal transform may be used instead of DCT. The residual image creation unit 100 and the residual image DCT unit 110 form a second residual image conversion unit.

  The image quality adjustment unit 120 adjusts the image quality and code data amount of the residual image converted into the frequency space pattern. A residual image IDCT unit 130 as a residual image inverse transform unit inversely transforms a frequency space pattern whose image quality has been adjusted to a residual image in real space by IDCT (Inverse Discrete Csine Transform). The reference image creation unit 150 creates a reference image using the residual image inversely transformed into the real space and the predicted image in the optimal prediction mode, and outputs this to the predicted image creation unit 140.

  As described above, according to the present embodiment, when optimal mode determination is performed based on a plurality of different prediction modes, optimal mode determination is performed by separately performing conversion processing for an original image and conversion processing for a predicted image. The amount of processing can be reduced.

  Here, FIG. 3 shows a configuration of a predictive coding apparatus 200 as a comparative example of the present embodiment. In addition, the same code | symbol is attached | subjected to the same element as the element shown by FIG. 1, and description is abbreviate | omitted. In the case of the predictive coding storage 200 of this comparative example, the residual image creating unit 100 calculates the residual between the original image and the predicted image and outputs the obtained residual image to the residual image Hadamard transform unit 210. The residual image Hadamard transform unit 210 performs a two-dimensional Hadamard transform on the residual image.

  In the predictive coding storage 200 of this comparative example, the two-dimensional Hadamard transform is performed a plurality of times for selecting the prediction mode. As a result, the processing amount is large and it is difficult to perform high-speed processing.

  As described above, in the case of the comparative example, a residual image is created for each prediction mode and the two-dimensional Hadamard transform is performed. In the present embodiment, the Hadamard transform of the original image and the predicted image are performed. The Hadamard transform is performed separately. Thereby, the calculation amount for calculating the residual image Hadamard transform pattern can be reduced.

  That is, if the predicted image is 0-dimensional or 1-dimensional, the Hadamard transform can be replaced with 0-dimensional and 1-dimensional images, respectively, and the processing amount can be reduced. Here, for example, a zero-dimensional prediction image means a DC prediction image in H.264 / AVC, and a one-dimensional prediction image means a vertical prediction image or a horizontal prediction image. . For reference, an example of the intra prediction mode in H.264 / AVC is shown in FIG.

  4A to 4C, FIG. 4A shows the intra 4 × 4 vertical prediction mode, FIG. 4B shows the intra 4 × 4 horizontal prediction mode, and FIG. 4C. Indicates an intra 4 × 4 DC prediction mode. As described above, the prediction mode is a prediction mode that can be frequency-converted to the zero dimension or the one dimension.

  The above-described embodiment is an example and does not limit the present invention. For example, the prediction mode may be selected from at least two prediction modes instead of three.

It is a block diagram which shows the structure of the prediction encoding apparatus by embodiment of this invention. It is explanatory drawing which shows the dimension of a prediction image. It is a block diagram which shows the structure of the prediction encoding apparatus by a comparative example. It is explanatory drawing which shows Intra 4x4 prediction mode in H.264 / AVC.

Explanation of symbols

10 Predictive coding apparatus 50 Original image Hadamard transform unit 60 Residual image Hadamard transform unit 70 SATD calculation unit 80 Mode selection unit 90 Predictive image Hadamard transform unit 100 Residual image creation unit 110 Residual image DCT unit 120 Image quality adjustment unit 130 Residual Difference image IDCT unit 140 Predictive image creation unit 150 Reference image creation unit

Claims (5)

A prediction image creating unit that creates a prediction image from a reference image based on the prediction mode selected from a plurality of different prediction modes;
An original image conversion unit for generating the original image having the first frequency conversion pattern by performing a first frequency conversion on the original image;
A prediction image conversion unit that generates the prediction image having the first frequency conversion pattern by performing the first frequency conversion on the prediction image;
A residual image having the first frequency conversion pattern is generated by calculating a difference between the original image having the first frequency conversion pattern and the predicted image having the first frequency conversion pattern. A first residual image conversion unit;
Based on the residual image having the first frequency conversion pattern, a mode determination unit that determines the prediction mode in which the code amount is small from the plurality of prediction modes;
A second residual image conversion unit that generates the residual image having a second frequency conversion pattern by performing a second frequency conversion on the residual image of the original image and the predicted image;
An image quality adjustment unit that performs image quality adjustment on the residual image having the second frequency conversion pattern;
The residual image for generating the residual image in the real space is obtained by performing inverse transformation of the second frequency transformation on the residual image having the second frequency transformation pattern that has undergone the image quality adjustment. A difference image inverse transform unit;
A predictive coding apparatus comprising: a reference image creating unit that creates the reference image using the residual image and the predicted image and outputs the reference image to the predicted image creating unit. The first frequency transform is a first orthogonal transform;
The predictive coding apparatus according to claim 1, wherein the second frequency transform is a second orthogonal transform using a transform matrix different from the first orthogonal transform. The first frequency transform is a Hadamard transform;
The predictive coding apparatus according to claim 1, wherein the second frequency transform is DCT. The prediction encoding apparatus according to claim 1, wherein the prediction mode is a prediction mode capable of frequency conversion to 0 dimension or 1 dimension. The prediction encoding apparatus according to claim 1, wherein the prediction mode is any one of a vertical prediction mode, a horizontal prediction mode, and a DC prediction mode.

Images