JP2005064638A - Method for compressing video signal - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for compressing a video signal capable of further applying lossless compression to the video signal subjected to lossy compression . <P>SOLUTION: Each block of compression frames configuring video signals subjected to lossy (irreversible) compression is used for a coding object block, and a recording form of the coding object block is revised in response to the number of coincident bytes of a reference block with the highest correlation by collating the coding object block with three reference blocks, an immediately preceding block, an immediately preceding line block, and an immediately preceding frame block. For example, when the coding object block is 9F (14, 2), the immediately preceding block is 9F (14, 2), the immediately preceding line block is 9F (15, 1), and the immediately preceding frame block becomes a block at the same position of the immediately preceding frame. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to storage and transmission of commercial or consumer video camera images, storage and transmission of master material in non-linear video production in which video images are captured and produced by a computer, medical video images in electronic medical records and telemedicine The present invention relates to a data compression technique suitable for storage and transmission.

  A video that is a moving image has a large amount of data. Therefore, if it is handled uncompressed, for example, the NTSC system has 30 M (mega) bytes per second, and the high vision system has 150 Mbytes. In other words, NTSC can store only 2.6 minutes and HDTV can store only 30 seconds per DVD. Therefore, conventionally, compression of video signals has been attempted by various methods.

  Most preferred is a lossless (reversible) compression method, which can be restored to its original state by decoding (see, for example, Patent Document 1). However, the lossless type has a limit of about 1/2 of the data compression rate. Therefore, the compression of the video signal cannot be completely restored to the original state when it is decoded, but a lossy (irreversible) type capable of reducing the compression rate to 1/5 to 1/10 is used. Yes.

  In particular, as digital video data recorded in a current VTR, a lossy compression method such as a business digital beta cam, a business DV cam, and a consumer DV is used. In both of these, the amount of data is reduced by performing compression by discrete cosine transform for each frame. Here, the compression method of the commercial DV cam and the consumer DV will be specifically described.

  First, FIG. 6A shows a schematic representation of an original video signal in a state where a video is photographed with a TV camera or the like and digitized. In FIG. 6, the left-right direction is a time-series direction, and the time advances as it goes to the right. The original video signal shown in FIG. 6A is composed of three planes of R, G, and B in which one frame is the three primary colors of light. Here, it is assumed that 30 frames per second and 8 bits are assigned to each color (plane) of each pixel. For example, a digital video signal recorded for 10 seconds becomes a frame group composed of 300 frames and 900 planes. The number of pixels in each frame is 720 × 480.

In the DV format, the digital video signal as described above is processed in units of frames as follows. First, the pixel signals of the R, G, and B planes are converted into Y, U, and V using the following [Equation 1].
[Formula 1]
Y = 0.299R + 0.587G + 0.114B
U = -0.1684R-0.3316G + 0.5B
V = 0.5R-0.4187G-0.0813B
Since the above [Formula 1] involves a decimal point operation, the original values of R, G, and B cannot be restored from Y, U, and V converted to integer values. In this state, the lossy type is basically used. become. Further, for the U plane and the V plane, the pixels are thinned out in half in the vertical direction and the horizontal direction. For example, in the case of an image of 720 × 480 pixels, an image of 360 × 240 pixels is obtained, and the number of pixels is reduced to ¼. The state of the video signal at this time is shown in FIG. Subsequently, in units of 8 × 8 pixels, one unit of the U plane and one unit of the V plane are allocated to one unit of the Y plane to form a block composed of a total of six units of pixels. Since one pixel is 1 byte, the data amount of this block is 384 bytes. Subsequently, one block is compressed into 80-byte data by performing DCT transform (discrete cosine transform), quantization, and variable length coding on this block.

When blocks are defined as described above, one original frame is represented by 45 × 30 blocks. In the DV format, a compressed block is defined as “9F”, and one header “7F” is added every 15 blocks. Furthermore, this one header “7F” and the following 15 blocks are set as one set, and a header composed of “1F”, “3F”, and “5F” is added every 9 sets. And configure the line. Then, the original frame is composed of 10 lines. This line is along the scanning line direction of the original image. The resulting compressed frame structure is shown in FIG. By converting to the DV format as described above, the video data is compressed from 1036800 bytes to 120,000 bytes. In the DV system, calculation processing between planes in the same frame is performed, but calculation between frames is not performed. For this reason, the concept of a plane is eliminated, but when there are n frames as shown in FIG. 6, n compressed frames as shown in FIG. 7 are created. The frame numbers F1 to Fn are also taken over as they are.
JP 11-164303 A

  However, even if the video signal is compressed by a lossy type such as DV, the amount of data is still large. For this reason, when data is transmitted, it may be necessary to perform further compression due to restrictions on a medium or a communication line. In this case, there is a desire to reduce the amount of data, but not to delete any more data from the video signal once lossy compressed. However, since the lossy compressed video signal has a unique structure as described above, the conventional lossless compression method adapted to the uncompressed video signal cannot be applied.

  Therefore, an object of the present invention is to provide a video signal compression method capable of further lossless compression of a video signal subjected to lossy compression.

  In order to solve the above problems, in the present invention, information is reproduced so that a compressed frame can be reproduced for a video signal composed of compressed frames in which each successive frame in time series is compressed by a lossy (irreversible) method. As a method of compressing the quantity, each pixel block obtained by blocking pixels in the compressed frame is set as an encoding target block, and immediately before the pixel block positioned immediately before in the scanning line direction with respect to each encoding target block. Compared with the three reference blocks of the block, the immediately preceding line block which is the pixel block located immediately before in the direction perpendicular to the scanning line direction, and the immediately preceding frame block which is the pixel block located at the same position of the immediately preceding frame, the most correlated The reference block selection stage for selecting a reference block with a high value, the selected reference block and the encoding target block As a correlation, a correlation determination stage that discriminates from three types of complete match, high correlation degree, and low correlation degree, and when the correlation is complete match as the output data of the encoding target block, the selected reference block If the correlation is large, the identification ID of the selected reference block, and the mask data indicating the position of the selected reference block and the matching data in the encoding target block in the block And a block encoding step that outputs data in the encoding target block that does not match the reference block, and outputs matching data and non-matching data in the encoding target block when the correlation is low. It is characterized by that.

  According to the present invention, with respect to the encoding target block in the compressed frame, three blocks of the scanning line direction, the previous block, the previous line block in the direction orthogonal to the scanning line, and the same block of the previous frame, that is, the image Three blocks adjacent to each other in the two directions and the time-series direction in the plane are compared as reference blocks, the reference block having the highest correlation is selected, and encoding is performed in different formats according to the degree of correlation with the reference block. Since it did in this way, compression is possible, without destroying the relationship between the blocks in a compression frame. That is, it becomes possible to further losslessly compress the video signal subjected to lossy compression.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
(Structure of lossy compressed video signal)
In the present embodiment, a case where processing is performed on a DV format video signal will be described as an example. As shown in FIG. 7, the information of the original 1 frame 3 plane (R, G, B) is converted into one compressed frame in the DV format. The data format is as described above, but considering the positional relationship with the original frame, one line in the compressed frame corresponds to three lines of the block. Therefore, here, for convenience of explanation, description will be made using a compressed frame moved to a state corresponding to the position of the original block as shown in FIG. FIG. 7 and FIG. 1 only show the same configuration except that the position of the block is changed.
(Compression method of the present invention)
Next, an outline of a video signal compression method according to the present invention will be described. The compression method of the present invention is executed by a computer and a dedicated software program installed in the computer. FIG. 2 is a flowchart showing an outline of a video signal compression method according to the present invention. First, a lossy-compressed video signal as shown in FIG. 1 is read by a device for lossless compression (computer equipped with dedicated software). Then, the device for lossless compression starts processing. First, among the frames of the lossy compressed video signal, the pixel blocks are sequentially read to be the encoding target blocks, and each encoding target block is compared with the immediately preceding block which is the immediately preceding pixel block (step S1). Specifically, the encoding target block 9F (i, j) is collated with the immediately preceding block 9F (i-1, j). For example, in the case of 9F (1, 1) shown in FIG. 1, since the immediately preceding block does not exist, collation is not performed. However, in the case of 9F (2, 1), the immediately preceding block 9F (1, 1) Are collated. As DV format data, the block immediately before 9F (1,2) is 9F (45,1), but as shown in FIG. 1, the block 9F (1,2) is the left end of the original image. Since the block 9F (45, 1) is a block located at the right end of the original image, it is considered that the correlation is low. Therefore, in such a case, collation with the immediately preceding block is not performed. In other words, when i = 1, collation with the immediately preceding block is performed. Collation is performed in units of bytes. Since one block is 80 bytes, collation is performed 80 times. Of these, how many bytes match is temporarily recorded in the memory. For example, if 9F (15, 2) matches 50 bytes with the immediately preceding block 9F (14, 2), collation data as shown in FIG. 3A is recorded.

  Next, collation with the immediately preceding line block which is a pixel block located immediately before in the direction perpendicular to the scanning line direction is performed (step S2). Specifically, the processing target block 9F (i, j) is collated with the immediately preceding line block 9F (i, j-1). In the case of 9F (1, 1) to 9F (45, 1) shown in FIG. 1, there is no immediately preceding line block, so no matching is performed, but in the case of 9F (15, 2), for example, the immediately preceding line A check with block 9F (15, 1) is performed. That is, when j is not 1, collation with the previous line block is performed. The matching method in step S2 is performed in units of bytes as in step S1, and the number of matched bytes is temporarily recorded. For example, if 9F (15, 2) matches 38 bytes with the immediately preceding line block 9F (15, 1), collation data as shown in FIG. 3B is recorded.

  Next, collation with the immediately preceding frame block existing at the same location of the immediately preceding frame is performed (step S3). All the frames constituting the lossy compressed video signal have the same configuration, as shown in FIG. Therefore, there is always a pixel block located at the same location as the encoding target block in the immediately preceding frame. Accordingly, in step S3, unlike in steps S1 and S2, collation is performed for all the encoding target blocks without exception. For example, in the case of 9F (15, 2) in the frame Fn, collation with 9F (15, 2) of the immediately preceding frame Fn−1 is performed. The collation itself is performed in units of bytes as in Steps S1 and S2, and the number of matched bytes is temporarily recorded. For example, if 9F (15, 2) matches 46 bytes with 9F (15, 2) of the immediately preceding frame Fn-1, collation data as shown in FIG. 3C is recorded.

  After performing the above-mentioned three collations of Step S1 to Step S3 for a certain encoding target block, the reference block having the highest correlation is selected from the collated reference blocks (Step S4). Specifically, the reference block having the maximum number of matching bytes temporarily recorded in the memory is selected. For example, in the example of FIG. 3C, since the number of matching bytes with 9F (14, 2) is the maximum, 9F (14, 2) is selected as the reference block. Here, the subsequent processing differs depending on the number of matching bytes of the selected reference block.

  For example, when the number of matching bytes with the selected reference block is 80 bytes, that is, when the data of the block to be encoded and the selected reference block are exactly the same, the mode “2” is output and the reference block Is output in association with the identification ID of the encoding target block (step S5). For example, when 9F (16, 3) completely matches the previous pixel block, information as shown in FIG. 4A is recorded. The identification ID may be any information as long as the information can identify the block. For example, as in 9F (14, 2) shown in FIG. 4, the order in the scanning line direction and the direction orthogonal to the scanning line may be used as the identification ID. Since the position of is known in the same way as the encoding target block, it can be specified by recording information indicating that it is the immediately preceding frame. The mode indicates an encoding pattern, and can be processed at a high speed by referring to it at the time of decoding.

  When the number of matching bytes is 20 bytes or more and less than 80 bytes (79 bytes or less), mode “1” is output, and position information of bytes that match the reference block in the encoding target block is output, and bytes that do not match The data itself is output (step S6). For example, when 50 bytes match with the reference block 9F (14, 2) for 9F (15, 2), information as shown in FIG. 4B is recorded.

  10 bytes are allocated for recording the mask data. Also, 10 bytes are allocated as a reserved area for recording the identification ID of the reference block and recording necessary information. Therefore, if the number of unmatched bytes is 60 bytes or more, the total data amount is 80 bytes or more. Therefore, if the number of matching bytes is less than 20 bytes (19 bytes or less), the mode “0” is output and the data of the encoding target block is output as it is without any change (step S7). For example, if the matching byte count of the reference block is less than 20 bytes for 9F (1, 5), information as shown in FIG. 4C is recorded.

  The above processing is performed on all the encoding target blocks of all the frames, and all the lossy compressed video signals are encoded. All the encoding target blocks are recorded in one of the formats (a) to (c) of FIG. The encoded information is recorded on an arbitrary recording medium in a format corresponding to the recording medium. Further, it may be transmitted via a network as it is.

In the above processing, only the pixel block with the most compression effect is processed as the encoding target block. This is because, in the DV format, 9F which is a pixel block occupies the majority of the data amount, so if only the pixel block is compressed, a considerable amount of data is reduced. However, the data amount may be further reduced by compressing blocks other than the pixel block. In this case, the control blocks “1F”, “3F”, and “5F” are not collated with the immediately preceding block in step S1, but are collated with the immediately preceding line block in step S2 and with the immediately preceding frame block in step S3. Perform verification. In step S2, the encoding target block 1F (j) is collated with the immediately preceding line block 1F (j-1), and 3F (i, j) and 5F (i, j) are respectively compared with the immediately preceding line. Collation with blocks 3F (i, j-1) and 5F (i, j-1) is performed. For the control block “7F”, three collations of Step S1 to Step S3 are performed in the same manner as the block “9F”.
(Decryption method)
Next, a method for decoding and decompressing code data compressed by the compression method will be described. Decoding is performed by a decoding device (comprising a computer and a dedicated software program installed in the computer). An outline of the decoding method is shown in the flowchart of FIG.

  The flowchart in FIG. 5 shows processing for each coding block. The encoded block is sequentially read from the head of each frame into the decoding device and processed. First, the encoding mode of the read encoding block is recognized (step S11). Specifically, it recognizes whether the mode of the coding block is 0, 1, or 2. In the case of mode “0”, that is, when data as shown in FIG. 4C is recorded, the data of the original encoding target block is recorded as it is, so that the encoded block read Is output as it is (step S12). In the case of mode “2”, that is, when data as shown in FIG. 4A is recorded, only the reference block ID is recorded, so the reference block specified by the corresponding identification ID Are output as they are (step S13). In the case of mode “0”, that is, when data as shown in FIG. 4B is recorded, the ID, mask data, and mismatched byte data of the reference block are recorded. In the reference block specified in (2), the byte data at the position specified by the mask data is extracted, and the extracted byte data and the mismatched byte data are integrated to restore the encoding target block (step S14).

  The block to be encoded is restored by repeatedly performing the processes in steps S11 to S14. As a result, data in the DV format as shown in FIG. 1 is restored.

It is a figure which shows the structure of the compression frame which comprises the video signal made into compression object by this invention. It is a flowchart which shows the outline | summary of the compression method of the video signal which concerns on this invention. It is a figure which shows the collation data temporarily recorded on a memory. It is a figure which shows the coding data obtained by this invention. It is a flowchart which shows the outline | summary of the process which decodes coding data. It is a figure which shows the uncompressed original video signal and the video signal obtained by the intermediate process of lossy compression. It is a figure which shows the structure of the compression frame of DV system.

Claims (4)

A method of compressing the amount of information so that the compressed frame can be reproduced with respect to a video signal composed of compressed frames obtained by compressing time-series continuous frames by a lossy method,
In the compressed frame, each pixel block obtained by blocking pixels is set as an encoding target block, and with respect to each encoding target block, the immediately preceding block which is a pixel block positioned immediately before in the scanning line direction, the scanning line direction, and The reference block with the highest correlation is selected by comparing with the three reference blocks of the previous line block, which is the pixel block located immediately before in the vertical direction, and the previous frame block, which is the pixel block located at the same position in the previous frame. A reference block selection stage to perform,
Correlation determination step for determining the correlation between the selected reference block and the encoding target block from three types of perfect match, high correlation degree, and low correlation degree;
As the output data of the block to be encoded, when the correlation is a perfect match, the identification ID of the selected reference block is output, and when the correlation is high in correlation, the selected reference block Output identification ID, mask data indicating the position of the selected reference block and the matching data in the encoding target block in the block, and data in the encoding target block that does not match the reference block, A block encoding step for outputting matched data and mismatched data in the encoding target block when the correlation is low, and
A method for compressing a video signal, comprising:   In the reference block selection step, for each control block other than the pixel block included in the video signal, the immediately preceding line control block which is a control block positioned immediately before in the direction perpendicular to the scanning line direction, and the same portion of the immediately preceding frame 2. The video signal compression according to claim 1, wherein the reference block having the highest correlation is further selected by collating with two reference blocks of the immediately preceding frame control block which is a control block located at Method.   The video signal is compressed by the DV or DVCAM system, and the pixel block is 80-byte fixed length data, which is equivalent to 6 code units of discrete cosine transform composed of 8 × 8 pixels in the original image before compression. The video signal compression method according to claim 1, wherein the video signal compression method is obtained by processing 1350 code units per NTSC specification frame. The recording medium of the video signal which recorded the code | cord | chord data obtained by any of the method of Claim 1 to 3 with respect to the given video signal.

Images