JPH09186995A - Data compression method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance the reduction rate of transmitted data by using variable length code tables different from each other with high code gain depending on the presence of an object in a unit block being divisions of each macro. SOLUTION: An object is in existence in all unit blocks Y1, Y2, Y3, Y4 of four macro blocks MB1, MB2, MB3, MB4. In the block MB2, the object is in existence in all unit blocks Y1-Y4, and in the block MB3, the object is in existence only in a 2nd unit block Y2, and the object is in existence only in 1st and end blocks Y1, Y2 in the block MB4. Since the object is selectively in existence in each unit block in each macro block, the coding gain is enhanced by applying a different variable length code table to each case and the data reduction efficiency is increased.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data compression method for converting video data in a frame into variable length coded data for transmission.

[0002]

2. Description of the Related Art In recent years, videophones, video conferences, etc.
H.264 provided by using an SDN (Integrated Service Digital Network). H. 261 and H.261 for videophone services using the public switched telephone network (PSTN). 263 has been proposed as a technical standard for the global standardization of image compression technology.

H.264, which provides a compression technique for ultra-low-speed moving images such as a videophone. 261 and H.E. In H.263, it is known that the quality of the image is deteriorated, and improvement for enhancing the image quality is required. On the other hand, the moving picture-related compression technique is MPEG1 (Motion Picture Expert Group) for DSM (Digital Storage Media).
p: Moving image expert group), DSM, HDTV
(High Density Television), A
MPEG2 for TV (Advanced Television) is standardized.

These H. 261, H .; 263, MPE
Block-based coding methods are used in G1 and MPEG2 compression techniques, but when these coding methods are applied to ultra-low-speed transmission, the image quality is noticeably degraded. For this reason, a completely new system, which is different from the existing system, has been proposed as MPEG4, and standardization is progressing to block-based coding in which the object concept is introduced.

Of these, the object-centered coding method (Object O
Riented Coding is based on an object in an area that moves between two images, not on a block, so that the entire image is displayed in the background or uncovered area.
covered region), a region where motion compensation is possible (MC-Regi
on) and a region in which motion compensation is not possible (MF-Region). Here, the motion compensable object (MC-object) refers to an object that is converted into a two-dimensional object in a three-dimensional space and moves according to a constant law such as horizontal movement, rotational movement, and linear deformation. An impossible object (MF-object) is an object to which this fixed law is not applied.

FIG. 7 is a view showing the relationship with one unit macroblock 33 in one frame grid 31. One frame grid 31 is divided into an array in the X-axis and Y-axis directions, and each divided becomes a unit macroblock 33. FIG. 7A shows a frame grid whose grid start position points are X-0 and Y-0.
FIG. 7B is an enlarged view of one unit macroblock, showing a state where one unit macroblock is divided into eight on the X axis and eight on the Y axis. Here, the divided elements are referred to as pixels 34. The unit macroblock 33 is called an N × M macroblock depending on the number of pixels. Therefore, the unit macroblock in FIG. 7B is an 8 × 8 macroblock.

A conventional pattern adaptive technique used as a technique for encoding the thus divided object will be considered with reference to FIG. Where the frame grid is 8
When the macroblock of × 8 is included and all the objects that cannot be motion compensated are filled in this macroblock, the efficiency is the same as that of the two-dimensional DCT (Discrete Cosine Transform). If not all the movement-compensated objects are filled in the block, the pattern portion of the corresponding movement-uncompensated object is subjected to the one-dimensional DCT on the X axis and further the one-dimensional DCT on the Y axis.

Hereinafter, the conventional pattern adaptive technique will be described in more detail. FIG. 8A shows an object that cannot be motion-compensated in an 8 × 8 macroblock (hatched portion).
Shows the existence of. In order to encode this object that cannot be motion-compensated by the pattern adaptive DCT, first, FIG.
As shown in (B), the upper boundary of the block is filled with an object that cannot be motion-compensated and is encoded in the vertical direction, that is, in the Y direction by the one-dimensional DCT.

The black circle in FIG. 8C is a one-dimensional D in the vertical direction.
It represents the discrete cosine transform (DC) value of CT.
In this state, Y-direction one-dimensional DCT is performed. When the one-dimensional DCT in the Y direction is performed and the state of FIG. 8D is obtained in this manner, the left boundary of the block is filled again and one-dimensional in the horizontal direction (X direction) as shown in FIG. 8E. DCT encoding is performed. When the X-direction one-dimensional DCT is completed, the X-direction and Y-direction one-dimensional pattern adaptive DCT is completed as shown in FIG. 8 (F).

A zigzag scan (ZIGZAG SCAN) is performed with the final form shown in FIG. 8F, and a moving object is scanned diagonally from the leftmost top block. MPEG-1, MPEG-using the block-centric coding described above
2, JPEG, H.M. 261, H .; The existing standard such as H.263 uses a macroblock MB having a form as shown in FIG. In this case, there are four blocks Y1, Y2, Y3, Y4 in one macroblock as shown in FIG. When encoding the four blocks existing in the macro block MB, each block Y1, Y2,
There are cases where data exists in Y3 and Y4 and cases where data does not exist. The existing standard uses a coded block pattern to indicate whether or not data exists in a block.

FIG. A variable length code table (variable lengt) for the code block pattern used in the H.263 standard.
h code table). This variable length code table is an index column showing 16 states, respectively, and CBPY (intra mode) showing a code block pattern (CBPY) in the intra (INTRA) mode.
Column, CBPY (inter mode) column showing the code block pattern (CBPY) in the inter mode, bit number column showing the bit length of the variable length coded data, and variable length coded data showing the variable length coded data. It is composed of columns.

Further, (1, 2, 3, 4) in the CBPY column is four unit blocks Y1, in which data may exist.
It corresponds to Y2, Y3, Y4, 1 for Y1 and 2 for Y
2, 3, 3 corresponds to Y3, and 4 corresponds to Y4. Therefore, for example, at index 1, CBPY (intra mode)
The value of (0,0,0,1) indicates that there is no data in blocks Y1, Y2 and Y3, and that there is data in block Y4. In this case, the variable length coded data has 5 bits and is represented by 00101. By using the variable-length code table in this way, it is possible to effectively indicate whether or not data is present in each macroblock, and transmit using the variable-length coded data obtained by conversion. The amount of data can be reduced, that is, data compression can be realized.

[0013]

However, in such a conventional data compression method by block-centered video signal coding, a code block pattern is used to indicate whether or not data is present in a macro block. , At this time, the variable length code table for the code block pattern is 1
However, there is a problem in that the efficiency of reducing the transmitted data is reduced because it is fixed.

That is, even if the number of objects is different from each other in one macroblock (for example, a small object exists only in a part of the macroblock or an object exists in the entire macroblock). Since the same variable length code table is applied regardless of the data reduction efficiency. The present invention is for solving the problem of the decrease in encoding efficiency that occurs when the conventional object-centered encoding is performed.
An object of the present invention is to provide a data compression method that increases the reduction efficiency of transmitted data.

[0015]

According to the data compression method of the present invention, one video frame is divided into macroblocks having a plurality of unit blocks, and whether or not video data exists in the unit blocks in each macroblock. In the data compression method, a code block pattern (CBPY) representing that is created, and a variable length code table is applied to the code block pattern (CBPY) to obtain variable length coded data corresponding to the code block pattern. Different variable length code tables are applied according to the number of unit blocks in which the video data may exist in the macro block.

[0016]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. A system to which the present invention is applied is a general video encoding system shown in FIG. 5, and a video signal is input, and a VOP (Video Object Plane: VOP) that divides a video of one frame into objects according to objects.
Video object plane) forming unit 10 and encoding unit 20 that encodes pattern information of each object obtained from the VOP forming unit 10
And each of the encoders 21 and 2 in the encoding unit 20.
, 23, ..., And the multiplexing unit 30 that multiplexes the coded video signals. VOP
The forming unit 10 divides the image information of one frame input from the image signal input unit for each object, and the encoders 21, 22, 23, ... In the encoding unit 20 encode the pattern information for each object. Turn into.

Here, one encoder 21 in the encoder 20 is constructed as shown in FIG. A switch 21a for selecting the object pattern information, and a pattern information encoding unit 2 for encoding the pattern information selected by the switch 21a.
1c, a selection switch 2 for selecting the pattern information coded by the pattern information coding unit 21c and arbitrary pattern information
1d, a motion estimation unit 21e that estimates the motion of the pattern information selected by the switch 21a based on the reproduced previous pattern information or arbitrary pattern information, and the motion estimation unit 2
1e, a motion compensating unit 21f for compensating the motion of the pattern information according to the motion amount estimated by 1e, a subtractor 21g for subtracting the image compensated by the motion compensating unit 21f from the image obtained by the VOP forming unit 10, A video signal coding unit 21h for coding the video signal obtained from the video signal converter 21g, and an adder 21i for adding the video information obtained from the video signal coding unit 21h and the video signal motion-compensated by the motion compensation unit 21f. , V for reproducing VOP with the video signal obtained from the adder 21i
The OP reproduction unit 21j and a memory 21b that stores variable-length code tables for different block patterns according to the number of blocks in which an object exists in a macroblock are stored.

The encoder 21 configured as described above selects the object pattern information obtained from the VOP forming section 10 by the switch 21a, and the selected pattern information is the pattern of the corresponding object by the pattern information coding section 21c. It is encoded as information. Then, the selection switch 21d selectively transmits the coded pattern information and the VOP of arbitrary pattern information to the multiplexing unit 21k, and the motion estimation unit 21e causes the VOP formation unit 1 to operate.
The motion compensation unit 2 estimates the motion from the pattern information obtained by
1f compensates the motion according to the estimated motion amount of the object.

The motion-compensated image is subtracted by the object pattern information obtained from the VOP forming unit 10 by the subtracter 21g, and then multiplexed by the image signal encoding 21h as encoded signal information of the image signal. It is transmitted to the section 21k. At this time, the video signal encoding unit 21h selectively applies variable length code tables for different block patterns stored in the memory 21b according to the number of blocks in which an object is present in the macro block to apply code blocks. Generate a pattern.

The video signal coded by the video signal coding unit 21h is added to the motion-compensated video signal by the adder 21i, and then reproduced by the VOP reproducing unit 21j as a previous VOP and reproduced. The motion estimation unit 21e estimates the motion of the object pattern information obtained from the VOP formation unit 10 by the VOP, and transmits the motion information estimated by the result to the multiplexing unit 21k.

Thereafter, the multiplexing unit 21k multiplexes the coded pattern information, the motion information and the signal information which are respectively input and outputs them, and the multiplexed signal is sent to a decoder (not shown) via the buffer unit 21m. It is transmitted as a bit stream. Here, the core of the present invention is to selectively apply a large number of variable-length code tables stored in the memory 21b according to the number of existing objects in each unit block in the macroblock.

Next, an embodiment of the present invention for this purpose will be described. That is, as shown in FIG. 1, since the object P does not necessarily have a square shape, when the object P is divided into macroblocks MB, blocks in which the object P is included and blocks in which the object P is not included in the divided macroblock MB. You can

As shown in FIG. 1 (A), an object P represented as video information data in one frame divided into a large number of macroblocks MB is a portion existing as data in each divided macroblock, There are some parts that do not exist. FIG. 1B is an enlarged view of a portion of a region Q surrounded by a dotted line in FIG. The hatched portion in the figure indicates the object P. In FIG. 1B, four macro blocks MB1, MB2, MB3, MB4 are shown, and each macro block includes four unit blocks Y.
1, Y2, Y3, Y4. Therefore, it can be seen that the macroblock MB includes a unit block in which the object P is included in MB4 and a unit block in which the object P is not included. This is distinguished and shown in FIG.

That is, in the macro block 1 (MB1) shown in FIG. 2A, all the objects exist in the other unit blocks Y2, Y3, Y4 except the first unit block Y1, and (B) The macroblock 2 (MB2) in () has all the objects in all the unit blocks Y1, Y2, Y3, Y4, and the macroblock 3 (MB3) in (C) has the objects only in the second unit block Y2. The macro block 4 (MB4) of (D) exists and the first unit block Y1
And the object exists only in the second unit block Y2.

Since an object is selectively present in each unit block in a macroblock as described above, different variable length code tables are applied to each case to increase coding gain and data reduction efficiency. Increase.

That is, since the object exists only in the three unit blocks Y2, Y3 and Y4 in the macro block 1 (MB1), data reduction can be performed by applying the variable length code table as shown in Table 1. The coding gain can be increased.

[Table 4]

Since all the objects exist in the four unit blocks Y1, Y2, Y3 and Y4 in the macro block 2 (MB2), in this case, the existing variable length code table shown in FIG. Since the object exists in only one unit block Y2 in the block 3 (MB3), the variable length code table as shown in Table 3 is applied to the macro block 4 (MB3).
Since the object exists only in the two unit blocks Y1 and Y2 in 4), the variable length code table as shown in Table 2 is applied.

[Table 5]

[Table 6]

In the above-described embodiment, the case where each macro block is divided into four square unit blocks and different variable length code tables are applied has been described, but the number of divisions is not limited to this. Without
When other division numbers are adopted, different variable-length code tables having the highest coding gains can be similarly obtained by experiments according to the existence of the object in each unit block. In this way, by applying different variable length code tables (variable length code table with the best coding gain) depending on the presence / absence of an object in each unit block in a macroblock, the data reduction rate and the coding gain are Can be increased.

Here, the different variable length code tables shown in Tables 1, 2, and 3 are new variable length code tables obtained by experiments, and the terms "INTRA" and "INTER" are used here. It is defined as follows.
The characteristics of the object on the time axis are those that did not exist in the previous frame but appeared in the current frame (object type I) and extremely small objects (objects within 10 pixels; object type II). And a relatively large object (object with 10 pixels or more; object type III). Here, there is an object that was not present in the previous frame but appeared in the current frame. , The encoding type is INTR for very small objects less than 10 pixels in size
A mode (non-prediction mode) is decided, and INTER mode (prediction mode) when the object does not have the above characteristics
To decide.

[0030]

As described above, according to the present invention, different variable length code tables having the highest coding gains are applied according to the existence of an object in a unit block obtained by dividing each macro block. Therefore, there is an advantage that the reduction rate of transmitted data can be increased and the coding gain can also be increased.

[Brief description of the drawings]

FIG. 1 is an exemplary diagram in which an arbitrary object is divided into macroblock units.

FIG. 2 is a view showing a unit block in which an object exists in a macroblock.

FIG. 3 shows MPEG-1, MPEG-2, H.264. 261,
H. A macroblock MB including four unit blocks Y1, Y2, Y3, Y4 used in the existing standard such as H.263.
FIG.

FIG. A variable length coding table for the coat block pattern used in the H.263 standard.

FIG. 5 is a block configuration diagram of an object video encoding system to which the present invention is applied.

6 is a detailed configuration diagram of one encoder 21 in FIG.

FIG. 7 is an explanatory diagram of a general frame grid and a macro block.

FIG. 8 is an explanatory diagram of general pattern adaptive DCT conversion.

[Explanation of symbols]

 10 VOP forming unit 20 Encoding unit 30 Multiplexing unit MB1 to MB4 Macroblock Y1 to Y4 Unit block

 ─────────────────────────────────────────────────── ───Continued from the front page (72) Inventor Kim Jong-Ion, Ivory villa B-202, 47-20 Samseong-dong, Gangnam-gu, Hanseong-si, Republic of Korea (72) Inventor Chikatsu Fumi Obayashi, Aobayashi, Sung-dong, Hanseong-gu, Republic of Korea Part 1-104

Claims (5)

[Claims] 1. A video frame is divided into macroblocks having a plurality of unit blocks, and a code block pattern (CBPY) representing whether video data is present in the unit blocks in each macroblock is created. ,
In a data compression method for obtaining a variable length coded data corresponding to the code block pattern by applying a variable length code table to the code block pattern (CBPY), it is possible that the video data exists in each macro block. A data compression method, wherein different variable length code tables are applied according to the number of certain unit blocks. 2. The data compression method according to claim 1, wherein the division of the macroblock is performed by dividing each macroblock into four square unit blocks. 3. The variable length code table defined in Table 1 is applied when the video data may be present only in three unit blocks in one macroblock. The data compression method according to 1 or 2. [Table 1] 4. The variable length code table defined in Table 2 is applied when the video data may be present only in two unit blocks in one macroblock. The data compression method according to Item 1 or 2. [Table 2] 5. The variable length code table defined in Table 3 is applied when the video data may exist in only one unit block in one macroblock. The data compression method according to 1 or 2. [Table 3]