JPH05236448A - Compressed moving image code amount controller - Google Patents

Abstract

PURPOSE:To execute encoding so as not to considerably degrade picture quality even when overflow is generated by reducing codes in the low priority order in the group of the prescribed number of blocks to be turned to a constant rate in the case of overflow. CONSTITUTION:In the entire group composed of the prescribed number of blocks to be turned to the constant rate, the priority order of variable length codes is judged by a code priority order judging equipment 14. When overflow is generated by a data stream former 15, the codes in the low priority order are reduced so as to settle the code amount in the group of the prescribed number of blocks to be stored in a data stream buffer 6 in a recording capacity. Afterwards, a corrected code, synchronizing code and ID code or the like are added by a corrected code former 7, converted to a code suitable for recording by a channel encoder 8 and recorded onto the recording medium by a head 9.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a compressed moving image code amount control device for making constant a compressed code amount of a device for compressing moving image data by combining orthogonal transform such as discrete cosine transform and variable length code. ..

[0002]

2. Description of the Related Art In a device for compressing the data amount of image data and recording it on a recording medium or transmitting it through a transmission path, the image data is divided into small blocks and orthogonal transformation such as discrete cosine transformation is performed and variable A method of allocating a long code and compressing the amount of data is well known. In such a device, it is necessary to control the code amount after compression to a constant rate according to the recording rate or the transmission rate. However, since the variable length code is used, there is a problem that the actual code amount cannot be known until encoding. On the other hand, conventionally, Japanese Patent Laid-Open No. Hei 3
As described in -229570, there has been proposed a method of predicting the code amount by using statistics of orthogonal transform coefficients. However, since it is impossible to perform 100% perfect prediction, a code overflow occurs in some images. The prior art has not taken into consideration that a code that seriously affects the image quality may not be transmitted when an overflow occurs, resulting in a large deterioration of the image quality.

[0003]

In the prior art, no consideration was given to deterioration of image quality at the time of overflow.

It is an object of the present invention to provide a code amount control device that causes little deterioration in image quality when an overflow occurs.

[0005]

In order to achieve the above object, means for storing information of all variable-length codes in a predetermined number of blocks for constant rate conversion, and a priority order between these codes are determined. And means for eliminating the overflow by deleting the code having a lower priority when overflowing.

[0006]

Since a code having a low priority is missing at the time of overflow, deterioration of image quality can be suppressed.

[0007]

1 is a block diagram of a recording apparatus according to a first embodiment of the present invention. The signal input to the video data input terminal 1 passes through the delay circuit 2, the discrete cosine transformer 3, the quantizer 4, the variable length encoder 5, the data stream buffer 6, the correction code former 7 and the channel encoder 8. , To the head 9. The variable length encoder 5 is connected to the code information storage unit 10. Further, the signal input to the video data input terminal 1 is also supplied to the code amount predictor 11, and the output of the code amount predictor 11 is supplied to the quantization controller 12.
The output of the quantization controller 12 is supplied to the quantizer 4. A signal is supplied from the code information storage unit 10 to the code amount detector 13, and the output of the code amount detector 13 is also the quantization controller 12
Is being supplied to. The output of the code information storage unit 10 is also supplied to the code priority determining unit 14, and the code priority determining unit 1
The output of 4 is supplied to the data stream former 15, and the output of the data stream former 15 is supplied to the variable length encoder 5 and the data stream buffer 6.

Next, the operation will be described. In the present embodiment, the incoming video data is divided into small blocks on the screen, the discrete cosine transform is performed for each block, the obtained transform coefficient is quantized, and the variable length code is assigned to the result to obtain the data. Perform compression. At this time, the quantizer 4, the code amount predictor 11, and the quantization controller 1 are included in order to keep the code amount after compression within the recording rate determined by the recording format.
2. The constant rate conversion circuit 16 constituted by the code amount detector 13 controls so that the code amount after variable length coding becomes constant for each group of a predetermined number of blocks. The code information storage unit 10 stores information on variable length codes for a group of a predetermined number of blocks for which the constant rate control is performed. The code amount is controlled by controlling the quantization step in the quantizer 4. First, the code amount predictor 11 predicts the code amount generated by a method such as performing quantization in advance in several quantization steps. The quantization controller 12 sets the quantization step in actual encoding to an appropriate value based on this prediction result. The delay circuit 2 is for increasing the processing time in the code amount predictor 11. The code amount detector 13 detects the code amount of the actual encoding result,
Feedback is given to the setting of the quantization step due to the deviation from the target value.

Although a constant rate is achieved as described above, since the code amount cannot be predicted with 100% certainty, the code amount may exceed the recording rate.
Therefore, the code priority determination unit 14 determines the priority of the variable-length code in the entire group of a predetermined number of blocks for which the constant rate conversion is performed, and when the data stream former 15 overflows, the priority is determined. The number of codes having a low number is reduced so that the code amount of a predetermined number of blocks stored in the data stream buffer 6 fits within the recording capacity. After that, the correction code forming unit 7 adds a correction code, a synchronization code, an ID code, etc.,
The channel encoder 8 converts the code into a code suitable for recording, and the head 9 records on the recording medium.

FIG. 2 shows a detailed block diagram of the encoding unit. The output of the quantizer 4 is supplied to the 0-run measuring device 52 through the scan converter 51, and information is sent from the 0-run measuring device 52 to the code information storage unit 10. Code information storage unit 10
From the code generator 53 to the code order converter 141. Code generator 53 to code table 54
Information is supplied to the code length table 55 and the code length table 55.
Information is returned to the data stream buffer 6 from the code generator 53. The output of the quantizer 4 is also supplied to the DC component encoder 56, and the output of the DC component encoder 56 is also supplied to the data stream buffer 6. On the other hand, the code order converter 14
A signal is supplied from 1 to the code generation controller 151, and a signal is supplied from the code generation controller 151 to the code generator 53. The information from the code length table 55 is also supplied to the overflow detector 152, and the overflow detector 15
A signal is also supplied from 2 to the code generation controller 151. A signal is also supplied from the overflow detector 152 to the block end code adder 153, and a signal is supplied from the block end code adder 153 to the data stream buffer 6.

Next, the operation will be described. FIG. 3A shows the structure of the discrete cosine transform block. The sampling frequency is changed between the luminance signal and the color difference signal, and four luminance blocks and two color difference blocks form one group (hereinafter referred to as a macro block). In this embodiment, one block is composed of 8 × 8 pixels. For each block, the DC component of the quantized discrete cosine transform coefficient is encoded as it is with a fixed length, and is written in the data stream buffer 6. For the AC component, first, the scan converter 51 performs zigzag scanning and rearranges in order from low frequency to high frequency. FIG. 3B shows the state of zigzag scanning. In FIG. 3B, arrows indicate the order of zigzag scanning, and a coefficient denoted by e in the block is a coefficient having a value other than 0, and the other coefficients are 0. The 0-run measuring device 52 measures the length of continuous 0s (0-run length) along the zigzag scan until a non-zero coefficient appears. The variable length code has a one-to-one correspondence with a combination of a 0 run length and a coefficient value (hereinafter, this is referred to as an event). If it is 0 until the end of the block, a special block end code is associated.

In this embodiment, the unit of constant rate conversion is 90 blocks (15 macroblocks). Hereinafter, this is referred to as a compressed block. When the codes of the blocks for one compressed block are arranged, it becomes as shown in FIG. Code information storage unit 10
Stores 0 run lengths and coefficient values of all events for one compressed block. The code order converter 141 converts the order of events, such as the first event of each block, and then the second event of each block. 1
The order in the second event is, as shown in the column of the generated code amount in FIG. 5, first the first block of the first macroblock and then the first block of the second macroblock. Alternate for each macroblock. The code generation controller 151 controls the code generator 53 to generate a code according to the order. The code generator 53 generates a code using the code table 54 and the code length table 55. The data stream buffer 6 stores the generated code. The inside of the data stream buffer 6 is divided for each macroblock and is stored as shown in FIG. As described above, a data stream for each macroblock is formed.

When the code amount overflows as shown in FIG. 5, the overflow detector 152 detects this and stops writing the code immediately before the overflow.
The block end code adding unit 153 is used to add the block end code to the block to which the block end code has not been written yet to end the writing so that the amount of the code stored in the data stream buffer 6 does not overflow. At this time, of course, the overflow is detected by counting the code amount of the block end code added at the end. As described above, in the example of FIG. 5, the symbols after e78 are deleted.

If the last AC component has a coefficient which is not 0 and is coded, the end of the block is known from this code, so the block end code need not be written.

FIG. 6 shows an example of the error correction product block formed in the correction code forming unit 7. After the synchronization signal, an ID indicating which macroblock is attached is added, and then a data stream of a fixed length DC component and an AC component is continued. The data that does not fit is embedded after the data of the macroblock with a small code amount. Since the total code amount of the data stream of the 15 macroblocks stored in the data stream buffer 6 does not overflow and is within a fixed amount, it can be stored in an error correction product block having a predetermined size. The outer correction code is added to the code string viewed in the vertical direction, and the inner correction code is added to the code string viewed in the horizontal direction.

As described above, according to the present embodiment, by reducing the code having the low priority in the entire group of the predetermined number of blocks which are subjected to the constant rate at the time of overflow,
Even if an overflow occurs, it is possible to perform encoding without causing a significant deterioration in image quality.

In the embodiments described above, the priority order of the same code of each block is determined by the order of the macroblock and the block, but the priority order is determined by the code length of each code. You can In Figure 7,
The detailed block diagram of such an encoding part of the 2nd Example of this invention is shown.

Quantizer 4, scan converter 51, 0-run measuring unit 52, code information storage unit 10, code generator 53, code table 54, code length table 55, DC component encoder 5
6, the code sequence converter 141, the code generation controller 151, and the overflow detector 152 are exactly the same as those in the embodiment shown in FIG. The output of the DC component encoder 56 is supplied to the data stream buffer 6, and the output of the code generator 53 is supplied to the data stream buffer 6 through the code buffer 66. The output of the overflow detector 152 is supplied to the code selector 154, the output of the code selector 154 is supplied to the code buffer 66 and the block end code adder 153, and the signal is output from the block end code adder 153 to the data stream buffer 6. Is being supplied. The output of the code length table 55 is supplied to the code length storage unit 155, and the signal is supplied from the code length storage unit 155 to the code selector 154.

Next, the operation will be described. The DC component has a fixed length and is written in the data stream buffer 6 as in the embodiment of FIG. Codes are generated for the AC component in the same order as in the embodiment of FIG. 2, but the same code in each block is temporarily stored in the code buffer 66. When no overflow occurs, the code stored in the code buffer 66 is written in the data stream buffer 6 as it is. Code length storage unit 155
Stores the code length information of the same code stored in the code buffer 66. When the overflow detector 152 detects an overflow, the code selector 154 uses this code length information to select the code in order from the code with the smallest code length until just before the overflow. FIG. 8B shows an example of the stored contents of the code length storage unit 155 when the eighth code overflows as shown in FIG. In this example, the code selection order is as shown in FIG.
e18 is excluded. The code selector 154 writes only the selected code into the data stream buffer 6, and uses the block end code adder 153 to add the block end code to the block to which the block end code has not been written yet, and ends the writing.

As described above, according to the present embodiment, by determining the priority in the same code of each block at the time of overflow, more optimal code reduction can be performed. Can be further reduced.

In the present embodiment, the priority order between the same number of codes is determined by the code length, but it may be configured such that the one having the larger coefficient value has the priority, or the zigzag scan shown in FIG. 3B. Faster position along (low frequency)
May be prioritized.

FIG. 9 shows a block diagram of an encoding unit according to the third embodiment of the present invention. Quantizer 4, scan converter 51, 0
Run measuring instrument 52, code generator 53, code table 54,
The configurations of the DC component encoder 56 and the data stream buffer 6 are exactly the same as those of the embodiment shown in FIG. Information is sent from the 0-run measuring device 52 to the code information storage unit 10. In addition, the 0 run measuring instrument 52 to the code length table 55
Information is supplied to the code information storage unit 10 from the code length table 55. Code information storage unit 10
From the code generator 53 and the reduction code selector 142. On the other hand, a signal is supplied from the reduction code selector 142 to the code generation controller 151, and the code generation controller 1
A signal is supplied from 51 to the code generator 53.

Next, the operation will be described. Code information storage unit 10
Stores 0 run length and three pieces of information obtained by adding code length information to the coefficient value. A specific example of the code information is shown in FIG. In the figure, the coefficient is represented by a numeral and the code length is represented by a length (0 run length is omitted). Reduction code selector 142
Using these pieces of information, when the code amount overflows as shown in FIG. 10B, the code to be deleted is selected until it does not overflow.

The codes are selected in order from the code having the smallest coefficient value among the codes at the end of each block (excluding the block end code). Those with the same coefficient value are selected from those with a long code length (those with a large reduction effect). As described above, e18 and e26 are deleted in the example of FIG.

For the block from which the last code has been deleted, the second last code is a candidate for the next deletion. However, in order to prevent concentration of image quality deterioration due to code deletion in one block, when deleting the second and subsequent codes, the error due to deletion (defined by the sum of squares of the deleted coefficient values) is evaluated. , The code of the block with the smaller value is deleted. Therefore, in the example of FIG. 4 (b), it is up to e26 to be deleted, but if it is next deleted (regardless of the code length), it is not e25 but e.
32.

The code generation controller 151 controls so that all the variable length codes (including the block end code) not selected by the reduction code selector 142 are written in the data stream buffer 6.

As described above, according to the present embodiment, the code having a small coefficient value is deleted at the time of overflow, and the image quality deterioration due to the code deletion is not concentrated on one block, so that the image quality deterioration at the time of overflow is not noticeable. can do.

In this embodiment, the code to be deleted is selected from the last code, but the code to be deleted is selected from all the codes regardless of the code position, and the code after the deleted code is deleted. Even if there is a code with a coefficient that is small in the middle, the code may be deleted so as to minimize the error by adopting a configuration in which the code is extended by 0 run length.

FIG. 11 shows a block diagram of an encoding unit according to the fourth embodiment of the present invention. Quantizer 4, scan converter 5
The configurations of the 1,0 run measuring device 52, the code information storage unit 10, the code generator 53, the code table 54, the code length table 55, the DC component encoder 56, the code sequence converter 141, and the code generation controller 151 are as shown in FIG. This is exactly the same as the embodiment shown in FIG. A signal is supplied from the code generator 53 to the first data stream buffer 61 and the second data stream buffer 62 through the buffer switch 18. The output of the DC component encoder 56 is supplied to the first data stream buffer 61. The information from the code length table 55 is also supplied to the first buffer overflow detector 17, and the signal is supplied from the first buffer overflow detector 17 to the buffer switch 18. A signal is also supplied from the first buffer overflow detector 17 to the block end code adder 153, and the block end code adder 1
A signal is supplied from 53 to the first data stream buffer 61. The outputs of the first data stream buffer 61 and the second data stream buffer 62 are supplied to the correction code forming unit 7.

Next, the operation will be described. In this embodiment, one compressed block for constant rate division is divided into three compressed sub-blocks. The code capacity of one compressed sub-block corresponds to the average code amount of 5 macroblocks, that is, 30 blocks. Depending on the block-to-block variation, the code amount of 5 macroblocks may fall within the code capacity of the compressed sub-blocks, and the code amount may exceed. While the overflow does not occur, the writing to the first data stream buffer 61 is performed in the same order as the writing to the data stream buffer 6 in the embodiment shown in FIG. When the code amount of a certain compressed sub-block overflows, the first buffer overflow detector 17 detects this, the writing of the code is stopped immediately before the overflow, and the block end code adder 153 is used to detect the block end code. First data stream buffer 6 added
The writing to 1 is completed, and the subsequent code is written to the second data stream buffer 62 by the buffer switch 18.

FIG. 12 shows an example of the data streams stored in the first data stream buffer 61 and the second data stream buffer 62 when the compressed sub-block overflows. The correction code former 7 forms an error correction product block from these data streams. FIG. 13 shows an error correction product block for the example of FIG. The data stream stored in the first data stream buffer 61 is stored in one compressed sub-block, and the data stream stored in the second data stream buffer 62 is stored in another compressed sub-block having a free space.

As described above, since one compressed sub-block contains a code with a high priority, even if the data stream is reproduced and decoded in units of 5 macroblocks, a reproduced image with little deterioration in image quality is obtained. Can be obtained.
Therefore, in the case of providing a mode in which the unit of constant rate conversion is expanded to 15 macroblocks for a format in which constant rate conversion is performed in units of 5 macroblocks, and a mode for improving image quality is provided
Compatibility can be ensured.

[0033]

According to the present invention, a code having a low priority is eliminated from the entire group of a predetermined number of blocks for which constant rate conversion is performed at the time of overflow, so that even if an overflow occurs, the image quality is greatly deteriorated. The encoding can be performed without any.

Further, by forming the sub-blocks with the highest priority, compatibility with the mode in which constant rate conversion is performed in sub-block units can be ensured.

[Brief description of drawings]

FIG. 1 is a block diagram of a recording apparatus according to a first embodiment of the present invention.

FIG. 2 is a detailed block diagram of an encoding unit according to the first embodiment.

FIG. 3 is a configuration diagram of a discrete cosine transform block.

FIG. 4 is a diagram showing an example of a variable length code of each block and a data stream of each macro block.

FIG. 5 is a diagram showing how codes are reduced in the first embodiment.

FIG. 6 is a configuration diagram of an error correction product block in the first embodiment.

FIG. 7 is a block diagram of an encoding unit according to a second embodiment of the present invention.

FIG. 8 is a diagram showing how codes are reduced in the second embodiment.

FIG. 9 is a block diagram of an encoding unit according to a third embodiment of the present invention.

FIG. 10 is a diagram showing how codes are reduced in the third embodiment.

FIG. 11 is a block diagram of an encoding unit according to a fourth embodiment of the present invention.

FIG. 12 is a diagram showing the contents of a data stream buffer in the fourth embodiment.

FIG. 13 is a configuration diagram of an error correction product block in the fourth embodiment.

[Explanation of symbols]

 5 ... Variable-length encoder, 6 ... Data stream buffer, 7 ... Correction code former, 10 ... Code information storage unit, 14 ... Code priority determination unit, 15 ... Data stream former.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takeshi Ichige 292 Yoshida-cho, Totsuka-ku, Yokohama, Kanagawa Prefecture

Claims (5)

[Claims] 1. A predetermined method in a moving image compression method for compressing data by dividing moving image data into blocks, performing orthogonal transformation, quantizing the obtained transform coefficient, and assigning a variable length code to the result. In a compressed moving image code amount control device for making a compressed code amount of a compressed block consisting of a number of blocks constant, code information storage means for storing information on all variable length codes in the compressed block, Priority-determining means for prioritizing the variable-length codes, and reducing the variable-length codes having a lower priority by deleting the variable-length codes having a lower priority when the code amount after compression exceeds a predetermined code amount. A compressed moving image code amount control device comprising means. 2. The priority determining means according to claim 1, wherein the variable length code corresponding to the low frequency transform coefficient in the compression block is rearranged to the variable length code corresponding to the high frequency transform coefficient. A compressed moving image code amount control device, wherein the code reduction unit is a unit that discards the variable length codes after that when the connection of the rearranged variable length codes exceeds a predetermined code amount. 3. The compressed moving picture code amount control device according to claim 2, further comprising means for adding a block end code to the block in which the variable length code is discarded by the code reduction means. 4. The priority determining means according to claim 1 is a means for selecting a variable length code having the lowest priority among the compressed blocks, and the code reducing means is a variable selected until the code quantity falls within a predetermined code amount. A compressed moving image code amount control device characterized in that it is means for deleting a long code. 5. A predetermined method in a moving image compression method for compressing data by dividing moving image data into blocks, performing orthogonal transformation, quantizing the obtained transform coefficient, and assigning a variable length code to the result. In a compressed moving picture code amount control device for making a compressed code amount of a compressed block composed of a number of blocks constant, information on all variable length codes in a compressed sub-block obtained by dividing the compressed block into a plurality of pieces is stored. Code information storage means, means for prioritizing variable-length codes in the compressed sub-blocks, and a code string with codes up to the code amount evenly allocated to the compressed sub-blocks in order from the code with the highest priority. Codes that are formed and have a lower priority that overflows the evenly allocated code amount are combined with the codes of other compressed sub-blocks that do not reach the evenly allocated code amount. A compressed moving image code amount control device, comprising: a code string forming means for forming a code string.