JP3481111B2 - Video decoding device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture decoding apparatus, for example, a moving picture suitable for decoding a signal compression-coded by the MPEG system to obtain a reproduced image having a resolution lower than that of an original image. The present invention relates to an image decoding device.

[0002]

2. Description of the Related Art An MPEG (Moving Picture Expert Group) system has been known as an image coding system for compressing and coding image data in the field of digital TV and the like.

A typical MPEG system is MPEG.
1 and MPEG2. In MPEG1, only progressive scan (non-interlaced) images are handled, but M
With PEG2, not only progressive scanning images but also interlaced scanning images have come to be handled.

For these MPEG encodings, motion compensation prediction (temporal compression), DCT (spatial compression) and entropy encoding (variable length encoding) are adopted. MP
In EG encoding, first, 16 (horizontal pixel number) × 1
Predictive coding in the time axis direction (frame predictive coding in MPEG1, frame predictive coding or field predictive coding in MPEG2) is performed for each macroblock unit having a size of 6 (the number of pixels in the vertical direction). 3 of I picture, P picture, and B picture corresponding to the predictive coding method
There are different image types. In the following, frame predictive coding will be described as an example.

(1) I picture: This is a screen encoded only from the information in the frame and is generated without performing inter-frame prediction. All macroblock types in the I picture are in the frame. is only intraframe sign-reduction to encode the information.

(2) P picture: This is a screen created by performing prediction from an I or P picture. Generally, the macroblock type in a P picture is an intraframe code that is coded using only intraframe information. Encoding and forward inter-frame predictive coding that is predicted from a past reproduced image.

(3) B picture: A picture which can be formed by bidirectional prediction, and generally includes the following macroblock types. a. Frame marks Goka b to encode only intraframe information. Forward inter-frame predictive coding for prediction from past reproduced images c. Backward inter-frame predictive coding for prediction from the future d. Interpolative Inter-frame Prediction Coding by Both Pre- and Post-Prediction Here, interpolative inter-frame prediction means averaging two predictions of forward prediction and backward prediction between corresponding pixels.

In the MPEG encoder, the image data of the original image is 16 (horizontal pixel number) × 16 (vertical pixel number)
It is divided into macroblock units of size. For macroblock type macro blocks other than sign-of frame, inter-frame corresponding to the macro block type prediction is performed, the prediction error data is generated.

[0009] Macro image data of the block each unit (if the macroblock type is sign-of frame) or prediction error data (if the macroblock type is inter-frame prediction coding) is an 8 × 8 The image data of each sub-block is divided into four sub-blocks of a size, and a two-dimensional discrete cosine transform (D
CT: Discrete Cosine Transform) is performed based on Equation 5. That is, as shown in FIG. 5, u is calculated based on each data f (i, j) in the block of 8 × 8 size.
Each DCT (orthogonal transform) coefficient F (u, v) in v space (u: horizontal frequency, v: vertical frequency) is obtained.

[0010]

[Equation 5]

In MPEG1, the DCT has a frame D.
Although only the CT mode is used, in the frame structure of MPEG2, it is possible to switch between the frame DCT mode and the field DCT mode in macroblock units. However, in the field structure of MPEG2, the field DC
Only in T mode.

In the frame DCT mode, a 16 × 16 macroblock is divided into four, and the upper left 8 × 8 block, the upper right 8 × 8 block, the lower left 8 × 8 block, and the lower right 8 × 8 block are divided. DCT is performed for each block.

On the other hand, in the field DCT mode, 16
An 8 × 8 data group consisting of only odd lines in an 8 (horizontal direction pixel number) × 16 (vertical direction pixel number) block in the left half of a × 16 macroblock, and an 8 × 16 block in the left half 8 × 8 data group consisting of only even lines, 8 × 8 data group consisting of only odd lines in the right half 8 (horizontal pixel number) × 16 (vertical pixel number) block and right half 8 D for each data group of the 8 × 8 data group consisting of only even lines in the × 16 block
CT is performed.

The DCT coefficient obtained as described above is quantized to generate a quantized DCT coefficient. The quantized DCT coefficients are zigzag-scanned or alternate-scanned, arranged in one dimension, and coded by a variable-length encoder. The MPEG encoder outputs the variable-length code of the transform coefficient obtained by the variable-length encoder, the control information including the information indicating the macroblock type, and the variable-length code of the motion vector.

FIG. 4 is a block diagram showing the structure of the MPEG decoder.

The variable length code of the transform coefficient is sent to the variable length decoder 101. Control signals including the macroblock type are sent to the CPU 110. The variable length code of the motion vector is sent to the variable length decoder 109 to be decoded. The motion vector obtained by the variable length decoder 109 is sent to the first reference image memory 106 and the second reference image memory 107 as a control signal for controlling the cutout position of the reference image.

The variable length decoder 101 decodes the variable length code of the transform coefficient. The inverse quantizer 102 receives the transform coefficient (quantized DCT) obtained from the variable length decoder 101.
Coefficient) is inversely quantized and converted into a DCT coefficient.

The inverse DCT circuit 103 includes an inverse quantizer 102.
The DCT coefficient sequence generated in step 1 is returned to the DCT coefficient in 8 × 8 sub-block units, and 8 × 8 inverse DCT is performed based on the inverse transform equation shown in Equation 6. That is, as shown in FIG. 5, based on the 8 × 8 DCT coefficient F (u, v),
Data f (i, j) in 8 × 8 sub-block units are obtained. Further, the data f (i,
Based on j), reproduced image data or prediction error data in units of one macroblock is generated.

[0019]

[Equation 6]

In the prediction error data in macroblock units generated by the inverse DCT circuit 103, reference image data according to the macroblock type is added by the adder 104.
Are added to generate reproduced image data. The reference image data is added to the adder 104 via the switch 112.
Sent to. However, if the data output from the inverse DCT circuit 103 is the playback image data for the sign-frame, the reference image data is not added.

The image data in macroblock units obtained by the inverse DCT circuit 103 or the adder 104 is B
If the reproduced image data is for a picture, the reproduced image data is sent to the switch 113.

When the reproduced image data in macro block units obtained by the inverse DCT circuit 103 or the adder 104 is the reproduced image data for the I picture or P picture, the reproduced image data is the switch 11
1 is stored in the first reference image memory 106 or the second reference image memory 107. The switch 111 is
It is controlled by the CPU 110.

The averaging unit 108 includes the memories 106 and 107.
The reproduced image data read from is averaged to generate reference image data used in the inter-frame predictive coding.

The switch 112 is controlled by the CPU 110 as follows. When data output from the inverse DCT circuit 103 is the playback image data for the frame marks items, the common terminal of the switch 112 is switched to the ground terminal.

When the data output from the inverse DCT circuit 103 is the prediction error data for the forward inter-frame prediction code or the prediction error data for the backward inter-frame prediction code, the common terminal of the switch 112 is the first terminal. It is switched to select either the terminal to which the output of the 1-reference image memory 106 is sent or the terminal to which the output of the second reference-image memory 107 is sent. When the reference image is read from the reference image memories 106 and 107, the cut-out position of the reference image is controlled based on the motion vector from the variable length decoder 109.

When the data output from the inverse DCT circuit 103 is prediction error data for the interpolative interframe prediction code, the common terminal of the switch 112 is the averaging unit 1.
It is switched to select the terminal to which the 08 output is sent.

The switch 113 stores reproduced image data for the B picture sent from the adder 104, reproduced image data for the I picture or P picture stored in the reference image memory 106, and stored in the reference image memory 107. The CPU 110 controls the reproduced image data for the I picture or P picture to be output in the same order as the original image. The image data output from the decoder is given to the monitor device, and the original image is displayed on the display screen of the monitor device.

[0028]

The above-mentioned conventional MPEG decoder has a problem that a memory for two screens is required for a reference image and the memory capacity is large.

An object of the present invention is to provide a moving picture decoding device capable of reducing the memory capacity.

[0030]

A moving picture decoding apparatus according to the present invention is a high-frequency part of vertical frequency among orthogonal transform coefficients in block units of a predetermined size obtained from an input signal.
By leaving, and, for performing an inverse orthogonal transform using a high-frequency coefficient reducing circuit for reducing only removed to transform coefficients in half the coefficients parts transform coefficients reduced by the factor reduction circuit of the horizontal frequency, Horizontal pixels for each block
Number on the basis of the reduced reproduction image data or the inverse orthogonal transform circuit to obtain a time base prediction error data, the time axis prediction obtained by the inverse orthogonal transform circuit error data and a predetermined reference image data to 1/2, The number of horizontal pixels is halved
Adder for generating a reduced reproduction image data, 1/2 reproduced image data obtained by the adder in the vertical direction
Thinned and the horizontal and vertical directions of the vertical thinning circuit number of pixels to generate reproduction image data is reduced to 1/2, respectively, and the reference image in the reproduced image data obtained by said vertical decimation circuit includes one or more reference image memory for storing the reproduced image data required to generate the data, only the number of pixels in the horizontal direction on the basis of the reproduced image data stored in the memory for the reference image 1
/ The reference image data which is reduced 2 to is characterized in that it is produced.

[0031] The predetermined block unit, in the horizontal image <br/> prime M, the number of pixels in the vertical direction is a block unit of size of M × N of N, at the time of coding the original image When orthogonal transform and was based on equation 7, it is used to perform inverse orthogonal transformation on the basis of equation 8 as the inverse orthogonal transform circuit.

[0032]

[Equation 7]

[0033]

[Equation 8]

[0034] The predetermined block unit, in the horizontal direction of the image <br/> prime 8, a block unit of size of 8 × 8 vertical pixel number 8, at the time of coding the original image When orthogonal transform and was based on equation 9, it is used to perform inverse transform on the basis of equation 10 as the inverse orthogonal transform circuit.

[0035]

[Equation 9]

[0036]

[Equation 10]

[0037] The vertical decimation circuit, a horizontal line of the reproduced image data obtained by said adder, for example, 2
Vertical thinning is performed by thinning every two units.

[0038]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention applied to an MPEG decoder will be described below with reference to FIGS.

FIG. 1 shows the structure of the MPEG decoder.

The variable length code of the transform coefficient is sent to the variable length decoder 1. Control signals including the macroblock type are sent to the CPU 20. The variable length code of the motion vector is sent to the variable length decoder 10 to be decoded. The motion vector obtained by the variable length decoder 10 is sent to the vector value conversion circuit 11 and converted so that the horizontal and vertical magnitudes of the motion vector are each halved. The motion vector converted by the vector value conversion circuit 11 so that the magnitudes in the horizontal direction and the vertical direction are each halved is stored in the first reference image memory 7 and the second reference image memory 8 as the reference image. It is sent as a control signal for controlling the cutting position.

The variable length decoder 1 decodes the variable length code of the transform coefficient. The inverse quantizer 2 inversely quantizes the transform coefficient (quantized DCT coefficient) obtained from the variable length decoder 1 and transforms it into a DCT coefficient. As shown in FIG. 2A, the horizontal high frequency coefficient removing circuit (coefficient reducing circuit) 3 sets the DCT coefficient sequence generated by the inverse quantizer 2 to 8 (horizontal pixel number) × 8 (vertical pixel number). Number) of 8 × 8 DCT coefficients F (u, v) (where u =
0, 1, ... 7, v = 0, 1 ,. Number of DCT coefficients F of frequency direction u) × 8 (vertical frequency direction v)
(U, v) (where u = 0, 1, ... 3, v = 0, 1,
… Convert to 7).

The inverse DCT circuit 4 subjects the 4 × 8 number of DCT coefficients generated by the horizontal high frequency coefficient removing circuit 3 to the inverse DCT of 4 × 8 as shown in Expression 11, and the result shown in FIG. ), The original sub-block unit data is 1 in the horizontal direction.
Data f (i, j) consisting of 4 (horizontal pixel number) × 8 (vertical pixel number) data compressed to 1/2 (where i = 0, 1, ... 3, j = 0,1) , ... 7) is generated.

[0043]

[Equation 11]

Also, one macroblock of 8 × 16 compressed in the horizontal direction by 1/2 based on the image data corresponding to the unit of four sub-blocks constituting one macroblock thus obtained. The unit reproduction image data or prediction error data is generated. Therefore, the inverse DC
The amount of data in macroblock units obtained by the T circuit 4 is half the amount of image data in macroblock units of the original image.

In the prediction error data generated by the inverse DCT circuit 4 in units of 8 × 16 macroblocks compressed in the horizontal direction by 1/2, reference image data (horizontal Reference image data in units of 8 × 16 macroblocks compressed to ½) are added by the adder 5 to generate reproduced image data. The reference image data is added to the adder 5 via the switch 13.
Sent to. However, if the image data output from the inverse DCT circuit 4 is the reproduction image data to the sign-frame, the reference image data is not added.

Reproduced image data obtained by the inverse DCT circuit 4 or the adder 5 in units of 8 × 16 macroblocks compressed in half in the horizontal direction is sent to the vertical thinning circuit 21. In the reproduced image data of 8 × 16 macroblock units obtained by the inverse DCT circuit 4 or the adder 5, as shown in FIG. 3A, the horizontal lines of odd fields (shown by solid lines) and the horizontal lines of even fields are horizontal. Since lines (indicated by broken lines) alternately appear in the vertical direction, in order to include the horizontal lines of the odd fields and the horizontal lines of the even fields, two horizontal lines are provided as shown in FIG. 3B. Thinning is performed in units.

As a result, the reproduced image data in units of 8 × 16 macroblocks compressed in the horizontal direction to ½ is
8 horizontally and vertically compressed to 1/2
It is converted into reproduced image data in units of × 8 macroblocks. Therefore, the amount of image data in macroblock units obtained by the vertical thinning circuit 21 is ¼ of the amount of image data in macroblock units of the original image.

When the reproduced image data in macro block units obtained by the vertical thinning circuit 21 is the reproduced image data for the B picture, the reproduced image data is sent to the switch 14.

The reproduced image data in macro block units obtained by the vertical thinning circuit 21 is I picture or P picture.
In the case of reproduction image data for a picture, the reproduction image data is stored in the first reference image memory 7 or the second reference image memory 8 via the switch 12.
First reference image memory 7 or second reference image memory 8
The amount of image data stored in is 1/4 of the conventional amount. The switch 12 is controlled by the CPU 20.

The first vertical interpolation circuit 22 performs vertical interpolation on the reference image data in 8 × 8 macroblock units read from the first reference image memory 7,
That is, the horizontal lines thinned out by the vertical thinning circuit 21 are interpolated to generate reference image data in 8 × 16 macroblock units.

The second vertical interpolation circuit 23 performs vertical interpolation on the reference image data in 8 × 8 macroblock units read from the second reference image memory 7,
That is, the horizontal lines thinned out by the vertical thinning circuit 21 are interpolated to generate reference image data in 8 × 16 macroblock units.

The averaging unit 9 averages the image data read from the first vertical interpolation circuit 22 and the second vertical interpolation circuit 23, and is used for interpolation interframe predictive coding 8 ×.
Reference image data of 16 macroblock units is generated.

The switch 13 is controlled by the CPU 20 as follows. When data output from the inverse DCT circuit 4 is the reproduction image data to the sign-of-frame, the common terminal of the switch 13 is switched to the ground terminal.

When the data output from the inverse DCT circuit 4 is the prediction error data for the forward interframe prediction code or the prediction error data for the backward interframe prediction code, the common terminal of the switch 13 is 1
It is switched to select either the terminal to which the reference image data from the vertical interpolation circuit 22 is sent or the terminal to which the reference image data from the second vertical interpolation circuit 23 is sent.

When the reference image is read from the reference image memories 7 and 8, the clipping position is controlled based on the motion vector from the vector value conversion circuit 11. The size of the motion vector in the horizontal and vertical directions is converted to ½ by the vector value conversion circuit 11 because the image in macro block units sent from the vertical thinning circuit 21 to the reference image memories 7 and 8 is used. This is because the data is compressed in half in the horizontal and vertical directions.

When the data output from the inverse DCT circuit 4 is the prediction error data for the interpolative inter-frame prediction code, the common terminal of the switch 13 selects the terminal to which the output of the averaging unit 9 is sent. Is switched to.

The switch 14 is the reproduction image data for the B picture sent from the vertical thinning circuit 21 to the switch 14, the reproduction image data for the I picture or P picture stored in the reference image memory 7, and the reference image memory. The CPU 20 controls the reproduced image data for the I-picture or P-picture stored in 8 to be output in the same order as the original image. Switch 14
The image data output from the format conversion circuit 1
5, the format is converted so as to correspond to the number of horizontal and vertical scanning lines of the monitor device and then sent to the monitor device.

According to the above-described embodiment, the first reference image memory 7 and the second reference image memory 8 can have the conventional capacity of 1/4.

By the way, in the 8 × 8 transform coefficient for each sub-block obtained from the inverse quantizer 2, not only the high frequency part of the horizontal frequency but also the wide frequency part of the vertical frequency is removed, and both the horizontal and vertical frequencies are removed. It is conceivable to perform the 4 × 4 inverse transform using only the 4 × 4 transform coefficient in the low region to generate data compressed to ½ in the horizontal and vertical directions. With this configuration, the capacities of the first reference image memory 7 and the second reference image memory 8 can be reduced to 1/4 of the conventional capacity, as in the above embodiment.

However, when not only the high-frequency part of the horizontal frequency but also the high-frequency part of the vertical frequency is removed and the 4 × 4 inverse transform is performed, if the image to be decoded is an interlaced image. Has a problem in that the spatial axis and the time axis are mixed to cause image deterioration. In order to prevent such image deterioration from occurring, in the above-described embodiment, only the coefficient in the high frequency part of the horizontal frequency is removed, and the coefficient in the high frequency part of the vertical frequency is left, and the inverse transform of 4 × 8 is performed. After that, vertical thinning is performed.

[0061]

According to the present invention, it is possible to obtain a moving picture decoding apparatus capable of reducing the memory capacity. Also the horizontal frequency
Only the coefficient of the high frequency part of the number is removed, and the high frequency part of the vertical frequency is removed.
Since the inverse orthogonal transform is performed leaving the coefficient of
If the image you are trying to interlace is
Prevents image deterioration caused by mixing the spatial and temporal axes.
You can

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an MPEG decoder.

FIG. 2 is a schematic diagram showing the DCT coefficients after the high-frequency part of the horizontal spatial frequency is removed by the horizontal high-frequency coefficient removing circuit, and the data after being inversely transformed by the inverse DCT circuit.

FIG. 3 is a schematic diagram for explaining thinning processing by a vertical thinning circuit.

FIG. 4 is a block diagram showing a configuration of a conventional MPEG decoder.

FIG. 5 is a schematic diagram for explaining DCT performed by an MPEG encoder and inverse DCT performed by a conventional MPEG decoder.

[Explanation of symbols]

1 Variable length decoder 2 Inverse quantizer 3 Horizontal high frequency coefficient removal circuit 4 Inverse DCT circuit 5 adder 7 First reference image memory 8 Second reference image memory 9 Averaging unit 10 Variable length decoder 11 Vector value conversion circuit 12, 13, 14 switch 15 Format conversion circuit 20 CPU 21 Vertical thinning circuit 22 First Vertical Interpolation Circuit 23 Second Vertical Interpolation Circuit

─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Akihiko Yamashita 2-5-5 Keihan Hondori, Moriguchi-shi, Osaka Sanyo Electric Co., Ltd. (56) Reference JP-A-9-247673 (JP, A) JP Japanese Patent Laid-Open No. 3-48587 (JP, A) Japanese Patent Laid-Open No. 5-304657 (JP, A) Japanese Patent Laid-Open No. 6-225268 (JP, A) Japanese Patent Laid-Open No. 8-116539 (JP, A) Japanese Patent Laid-Open No. 8-186818 (JP , A) JP 8-205161 (JP, A) JP 9-247671 (JP, A) JP 9-252469 (JP, A) US Patent 5262854 (US, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68

Claims (4)

(57) [Claims] 1. A high-frequency part of a vertical frequency in an orthogonal transform coefficient of a block unit of a predetermined size obtained from an input signal.
Leaving the minute, and the coefficient reducing circuit for reducing by half the high-frequency removal to transform coefficients coefficients only part of the horizontal frequency, by performing inverse orthogonal transformation using the transformation coefficient that has been reduced by the factor reduction circuit , inverse orthogonal transform circuit for obtaining a reproduced image data or time axis prediction error data number of pixels in the horizontal direction for each block is reduced to 1/2, the time axis prediction error data and a predetermined obtained by the inverse orthogonal transform circuit Based on the reference image data of
Adder number of pixels to generate reproduction image data is reduced to 1/2, the reproduced image data obtained by the adder thinned to 1/2 in the vertical direction, the number of horizontal and vertical pixels <br /> vertical thinning circuit to generate the reproduced image data is reduced to 1/2, respectively, and reproduced image necessary for generating the reference image data of the reproduced image data obtained by said vertical decimation circuit includes one or more reference image memory for storing data, said <br/> reference only number of pixels in the horizontal direction is reduced to 1/2 on the basis of the reproduced image data stored in the memory for the reference image A moving picture decoding device for generating image data. Wherein said predetermined block unit, in <br/> number of pixels in the horizontal direction is M, a block of the size of the number of pixels in the vertical direction is N of M × N, the coding of the original image The moving picture decoding apparatus according to claim 1, wherein if the orthogonal transformation is performed based on Equation 1, the inverse orthogonal transformation circuit performs inverse orthogonal transformation based on Equation 2. [Equation 1] [Equation 2] Wherein the predetermined block unit, in the horizontal direction of <br/> number of pixels is 8, a block unit of size of 8 × 8 vertical pixel number 8, the encoding of the original image The moving picture decoding apparatus according to claim 1, wherein if the orthogonal transformation is performed based on Equation 3, the inverse orthogonal transformation circuit performs inverse transformation based on Equation 4. [Equation 3] [Equation 4] 4. The vertical thinning circuit thins the horizontal lines of the reproduced image data obtained by the adder in units of two every two lines in the vertical direction. 4. The moving picture decoding device according to any one of 3 and 3 .