JP2007053788A - Video data compression apparatus and method thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a video data compression apparatus and a method thereof for adjusting a produced bit amount in response to the complicatedness of a pattern of each part of video data, so as to improve the quality of the video image after compression as a whole. <P>SOLUTION: A real difficulty data calculation circuit 282 calculates real difficulty data Dj corresponding to a data amount after the compression on the basis of index data (ME residual, flatness, intra AC, and activity) indicating the complicatedness of a pattern of a video image. A parameter calculation circuit 286 adjusts a parameter Rj' in response to an occupancy factor or the like of a VBV buffer. A target data amount calculation circuit 284 multiplies a multiplier calculated on the basis of the real difficulty data Dj with the parameter Rj' adjusted by the parameter calculation circuit 286 to calculate a target data amount Tj indicating a target value of the data amount of the video image after the compression. Rate control is achieved by calculating a quantization index QIND from the target data amount Tj and using it for compression coding processing. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a video data compression apparatus and method for compressing and encoding uncompressed video data.

  GOP composed of I picture (intra coded picture), B picture (bi-directionally predictive coded picture) and P picture (predictive coded picture) by uncompressed digital video data by a method such as MPEG (moving picture experts group) When recording on a recording medium such as a magneto-optical disc (MO disc) by compressing and encoding in units of groups, the amount of compressed video data (bit amount) after compression encoding is set. It is necessary to keep the recording quality of the recording medium or less than the transmission capacity of the communication line while keeping the quality of the video after decompression decoding high.

  For this purpose, first, uncompressed video data is preliminarily compressed and encoded, and the amount of data after compression encoding is estimated (first pass). Next, the compression rate is adjusted based on the estimated amount of data and compressed. A compression encoding method (second pass) is adopted so that the amount of data after encoding is equal to or less than the recording capacity of the recording medium (hereinafter, such compression encoding method is also referred to as “two-pass encoding”).

  However, if compression encoding is performed by two-pass encoding, it is necessary to perform similar compression encoding processing twice on the same uncompressed video data, which takes time. In addition, since the final compressed video data cannot be calculated by a single compression encoding process, the captured video data cannot be directly compressed and recorded in real time (real time).

The present invention has been made in view of the above-described problems of the prior art, and a video data compression apparatus capable of compressing and encoding audio / video data below a predetermined amount of data without using two-pass encoding, and An object is to provide such a method.
Another object of the present invention is to provide a video data compression apparatus and method capable of compressing and encoding video data substantially in real time and obtaining a high-quality video after decompression decoding. To do.
The present invention also provides a video data compression apparatus and method capable of performing compression coding processing by estimating the amount of data after compression coding and adjusting the compression rate without using two-pass encoding. With the goal.

  In order to achieve the above object, a video data compression apparatus according to the present invention is a data that satisfies a condition that is determined based on a VBV buffer that buffers uncompressed video data of a moving image and video data after compression (compressed video data) A video data compression apparatus for compressing to a rate, difficulty level data calculating means for calculating difficulty level data indicating the complexity of the video of the uncompressed video data, and data of the compressed video data buffered in the VBV buffer A data amount allocating means for allocating a compressed data amount (allocated data amount) to a predetermined number of pictures of uncompressed video data based on the amount (occupied data amount) and the calculated difficulty data, and the calculated difficulty level Based on the data and the allocated data volume, the target value of the compressed data volume of the uncompressed video data Has a target value calculating means for calculating, said by the method prescribed compression uncompressed video data, and compression means the amount of compressed data is compressed so that the target value calculated in.

  Preferably, the difficulty level data calculating means calculates a data amount after compression of the uncompressed video data as the difficulty level data.

  Preferably, the data amount allocating means is configured such that the video of the uncompressed video data is complex based on the calculated difficulty data when the occupied data amount of the VBV buffer is equal to or greater than a predetermined threshold. The value of the allocated data amount is increased, and the target value calculation unit increases the target value as the video of the uncompressed video data is more complex, and the video of the uncompressed video data is simpler. The more the value is, the smaller the target value is.

  Preferably, the data amount allocating unit subtracts a reference value obtained by evenly distributing all the data amount given to the compressed video data to all the pictures from the data amount of the generated compressed video data picture. A difference value is calculated, and when the compression of the uncompressed video data is completed, the allocated data amount is calculated so that the sum of the calculated difference values becomes a value close to a negative value of zero.

  Preferably, the target amount calculating means calculates the target value by multiplying the calculated amount of assigned data by a value obtained by dividing the difficulty data of the latest picture by the sum of the difficulty data of a predetermined number of pictures.

  Preferably, the compression means compresses the uncompressed video data into a picture type sequence including a plurality of types of pictures (I picture, P picture and B picture or a combination thereof) in a predetermined order.

  Preferably, the difficulty level data calculating means includes, as the difficulty level data, an ME residual of a picture compressed into a P picture or a B picture, and a flatness, intra AC data and activity of a picture compressed into an I picture, or These combinations are calculated.

  Preferably, the data amount allocating unit adds, as the predetermined threshold value of the occupied data amount of the VBV buffer, an addition value corresponding to the data rate of the generated compressed video data to the latest I picture data amount, or A numerical value obtained by adding a fixed addition value is used.

  Preferably, the data amount allocating means determines whether or not the occupied data amount of the VBV buffer is equal to or greater than a predetermined threshold immediately after the compression means compresses the uncompressed video data into a P picture. Do.

  The video data compression apparatus according to the present invention converts uncompressed video data of a moving image into a picture type sequence including a plurality of types of pictures (I picture, P picture, and B picture) in a predetermined combination and order by, for example, MPEG. Compression-encoded, the amount of video data after compression-encoding (compressed video data) is kept below the recording capacity of the recording medium, etc., and satisfies the constraint on the VBV (video buffering verifier) buffer occupancy, Generate compressed video data with high quality video after decompression decoding.

  In the video data compression apparatus according to the present invention, the difficulty level data calculation means generates difficulty level data indicating the complexity (simpleness) of the video of the uncompressed video data. The difficulty data generated by the difficulty data calculation means includes, for example, ME residual, actual difficulty data Dj corresponding to the data amount of compressed video data obtained by pre-compressing non-compressed video data, or video flatness Newly defined flatness for indexing, flat AC, intra-AC, activity used to calculate the quantized value of macroblock (MQUANT) in TM5, and global compression Global complexity is used.

  The data amount allocating means is a data amount that allows a predetermined margin in the data amount of the latest I picture that is equal to or greater than a predetermined threshold in the VBV buffer so that an underflow does not occur in the VBV buffer that buffers the compressed video data. Only when the above compressed video data is buffered, the allocation data amount indicating the data amount that can be allocated to the compressed picture of the uncompressed video data is adjusted. Note that the determination of the data amount of the compressed video data buffered in the VBV buffer does not need to be performed every time the compression unit generates a picture of the compressed video data, and the data amount in the VBV buffer is used for this determination. It is preferable that the compression means that achieves the optimum value is performed immediately after the P picture is generated.

  The adjustment of the allocated data amount by the data amount allocating means is, for example, a reference value in which the maximum data amount (maximum data amount; for example, recording capacity of the recording medium) allowed as a whole for the compressed video data is evenly allocated to each picture. If the VBV buffer occupancy is more than a certain value (when underflow will not occur), the predetermined number (L) of pictures with a complicated video will be more than the reference value. Allocation data is allocated and allocation data smaller than the reference value is allocated to pictures with simple video. Finally, the amount of compressed video data does not exceed the maximum data amount, and the amount of compressed video data is almost the same. Make the maximum amount of data.

  It should be noted that when adjusting the amount of allocated data by the data amount allocating means, not only a condition for the underflow of the VBV buffer is provided, but naturally the data rate of the compressed video data does not cause an overflow in the VBV buffer. It is necessary to satisfy the conditions.

The target amount calculation means, for example, multiplies the assignment data calculated by the data amount assignment means by multiplying the value obtained by dividing the difficulty data of the latest picture by the total value of the difficulty data of the L pictures into the uncompressed video data. The target value of the data amount after compression for each of the pictures is calculated.
In this way, the maximum data amount can be effectively used by the data amount allocating unit and the target amount calculating unit, and compressed video data with a large amount of data is generated from a complex video picture that requires a large amount of data, A small amount of compressed video data is generated from a simple video picture that requires a small amount of data, and the quality of the compressed video data as a whole improves.

  The compression means compresses and encodes the non-compressed video data by, for example, the MPEG method so that the data amount after compression encoding is substantially the target value calculated by the target amount calculation means in a predetermined picture type sequence. .

  The video data compression method according to the present invention compresses uncompressed video data of a moving image to a data rate that satisfies a condition determined based on a VBV buffer that buffers the compressed video data (compressed video data). A compression method, after compressing a predetermined number of pictures of uncompressed video data based on the amount of data until the VBV buffer underflows and the complexity of the pictures of the pictures of the uncompressed video data A target amount of data after compression of the uncompressed video data based on the complexity of the picture of the picture of the uncompressed video data and the calculated allocated data amount Is calculated for each picture, and the uncompressed video data is calculated according to a predetermined compression method, and the amount of data after compression is calculated. To compress so as to.

The video data compression apparatus and method according to the present invention can compress and encode audio / video data below a predetermined data amount without using two-pass encoding.
Also, according to the video data compression apparatus and method according to the present invention, video data can be compression-encoded substantially in real time, and high-quality video can be obtained after decompression decoding.
Further, according to the video data compression apparatus and method according to the present invention, it is possible to perform compression coding processing by adjusting the compression rate by estimating the data amount after compression coding without using two-pass encoding. .

First Embodiment Hereinafter, a first embodiment of the present invention will be described.
When compression coding of video data such as MPEG, images with many high frequency components or graphics with high difficulty, such as graphics with a lot of movement, are generally susceptible to distortion caused by compression. Become. For this reason, video data with a high degree of difficulty must be compression-encoded at a low compression rate. For compressed video data obtained by compression-encoding data with a high degree of difficulty, compressed video of video data with a low degree of difficulty is used. It is necessary to allocate a larger amount of target data than data.

Thus, in order to adaptively allocate the target data amount to the difficulty level of the video data, the two-pass encoding method shown as the prior art is effective. However, the two-pass encoding method is not suitable for real-time compression encoding.
The simple two-pass encoding method shown as the first embodiment is made to solve the problems of the two-pass encoding method, and is compressed video data obtained by preliminarily compressing and encoding uncompressed video data. The difficulty level of the uncompressed video data is calculated from the difficulty level data, and the compression rate of the uncompressed video data delayed by a predetermined time by the FIFO memory is adaptively controlled based on the difficulty level calculated by the preliminary compression encoding. can do.

FIG. 1 is a diagram showing a configuration of a video data compression apparatus 1 according to the present invention.
As shown in FIG. 1, the video data compression apparatus 1 includes a compression encoding unit 10 and a host computer 20. The compression encoding unit 10 includes an encoder control unit 12, a motion estimator 14, a simple 2 The path processing unit 16 and the second encoder 18 are included, and the simple two-pass processing unit 16 includes a FIFO memory 160 and a first encoder 162.
With these components, the video data compression apparatus 1 realizes the above-described simple two-pass encoding for uncompressed video data VIN input from an external device (not shown) such as an editing device and a video tape recorder device. To do.

  In the video data compression apparatus 1, the host computer 20 controls the operation of each component of the video data compression apparatus 1. In addition, the host computer 20 determines the amount of compressed video data generated by pre-compressing the uncompressed video data VIN by the encoder 162 of the simple two-pass processing unit 16 and the direct current component (DC) of the video data after DCT processing. The component value and the DC component (AC component) power value are received via the control signal C16, and based on the received values, the degree of difficulty of the pattern of the compressed video data is calculated. Further, the host computer 20 assigns a target data amount Tj of the compressed video data generated by the encoder 18 for each picture based on the calculated difficulty level, and a quantization circuit 166 of the encoder 18 (FIG. 3). And the compression rate of the encoder 18 is adaptively controlled for each picture.

The encoder control unit 12 notifies the host computer 20 of the presence or absence of a picture of the uncompressed video data VIN, and further performs preprocessing for compression encoding for each picture of the uncompressed video data VIN. That is, the encoder control unit 12 rearranges the input uncompressed video data in the order of encoding, performs picture field conversion, and performs 3: 2 pull-down processing (movie) when the uncompressed video data VIN is movie video data. The video data of 24 frames / second is converted into video data of 30 frames / second and the redundancy is removed before compression encoding), and the like, and the FIFO memory 160 of the simple two-pass processing unit 16 is used as the video data S12. And output to the encoder 162.
The motion detector 14 detects a motion vector of the uncompressed video data and outputs it to the encoder control unit 12 and the encoders 162 and 18.

  In the simple 2-pass processing unit 16, the FIFO memory 160 delays the video data S12 input from the encoder control unit 12 by, for example, a time during which L (L is an integer) picture input of the uncompressed video data VIN, The delayed video data S16 is output to the encoder 18.

FIG. 2 is a diagram illustrating a configuration of the encoder 162 of the simple two-pass processing unit 16 illustrated in FIG.
For example, as shown in FIG. 2, the encoder 162 includes an adder circuit 164, a DCT circuit 166, a quantization circuit (Q) 168, a variable length coding circuit (VLC) 170, an inverse quantization circuit (IQ) 172, and an inverse DCT. (IDCT) A general video data compression encoder composed of an (IDCT) circuit 174, an adder circuit 176, and a motion compensation circuit 178, wherein the input video data S12 is compressed and encoded by the MPEG method, etc. The amount of data for each picture is output to the host computer 20.

The adder circuit 164 subtracts the output data of the adder circuit 176 from the video data S12 and outputs it to the DCT circuit 166.
The DCT circuit 166 performs discrete cosine transform (DCT) processing on the video data input from the adder circuit 164, for example, in units of macroblocks of 16 pixels × 16 pixels, and converts from time domain data to frequency domain data. It outputs to the quantization circuit 168. Further, the DCT circuit 166 outputs the DC component value and the AC component power value of the video data after DCT to the host computer 20.

The quantization circuit 168 quantizes the frequency domain data input from the DCT circuit 166 with a fixed quantization value Q, and outputs the quantized data to the variable length encoding circuit 170 and the inverse quantization circuit 172. .
The variable-length coding circuit 170 performs variable-length coding on the quantized data input from the quantization circuit 168, and the amount of compressed video data obtained as a result of the variable-length coding is hosted via the control signal C16. Output to the computer 20.
The inverse quantization circuit 172 inversely quantizes the quantized data input from the variable length encoding circuit 168 and outputs the inverse quantized data to the inverse DCT circuit 174.

The inverse DCT circuit 174 performs inverse DCT processing on the inversely quantized data input from the inverse quantization circuit 172 and outputs the result to the adder circuit 176.
The adder circuit 176 adds the output data of the motion compensation circuit 178 and the output data of the inverse DCT circuit 174 and outputs the result to the adder circuit 164 and the motion compensation circuit 178.
The motion compensation circuit 178 performs motion compensation processing on the output data of the addition circuit 176 based on the motion vector input from the motion detector 14 and outputs the result to the addition circuit 176.

FIG. 3 is a diagram showing a configuration of the encoder 18 shown in FIG.
As shown in FIG. 3, the encoder 18 has a configuration in which a quantization control circuit 180 is added to the encoder 162 shown in FIG. Based on the target data amount Tj set from the host computer 20, the encoder 18 performs motion compensation processing, DCT processing, quantum processing on the delayed video data S16 delayed by L pictures by the FIFO memory 160 based on the target data amount Tj set by the host computer 20. Compressed video data VOUT such as MPEG format is generated and output to an external device (not shown).

In the encoder 18, the quantization control circuit 180 sequentially monitors the data amount of the compressed video data VOUT output from the variable length quantization circuit 170, and is finally generated from the j-th picture of the delayed video data S16. The quantization value Qj set in the quantization circuit 168 is sequentially adjusted so that the data amount of the compressed video data approaches the target data amount Tj set from the host computer 20.
In addition to outputting the compressed video data VOUT to the outside, the variable length quantization circuit 170 outputs the actual data amount Sj of the compressed video data VOUT obtained by compressing and encoding the delayed video data S16 via the control signal C18. To the host computer 20.

Hereinafter, a simple two-pass encoding operation of the video data compression apparatus 1 in the first embodiment will be described.
4A to 4C are diagrams illustrating a simple two-pass encoding operation of the video data compression apparatus 1 according to the first embodiment.
The encoder control unit 12 performs pre-processing such as rearranging pictures in the encoding order by the encoder control unit 12 with respect to the uncompressed video data VIN input to the video data compression device 1, and is shown in FIG. As described above, the video data S12 is output to the FIFO memory 160 and the encoder 162.
It should be noted that the picture order rearrangement by the encoder control unit 12 causes the picture coding order shown in FIG. 4 and the like to be different from the display order after decompression decoding.

The FIFO memory 160 delays each picture of the input video data S12 by L pictures and outputs it to the encoder 18.
The encoder 162 preliminarily sequentially compresses and encodes the pictures of the input video data S12, and compresses and encodes the jth (j is an integer) picture, and the DCT process The DC component value and AC component power value of the subsequent video data are output to the host computer 20.

  For example, since the delayed video data S16 input to the encoder 18 is delayed by L pictures by the FIFO memory 160, as shown in FIG. 4B, the encoder 18 performs the j-th (j) of the delayed video data S16. Is an integer) -th picture (picture a in FIG. 4B), the encoder 162 encodes the (j + L) -th picture ahead of the j-th picture in the video data S12. The picture (picture b in FIG. 4B) is compression-encoded. Therefore, when the encoder 18 starts compression encoding of the jth picture of the delayed video data S16, the encoder 162 uses the jth to (j + L-1) th pictures (FIG. 4 (FIG. 4)). The compression encoding in the range c) of B) has been completed, and the actual difficulty data Dj, Dj + 1, Dj + 2,..., Dj + L-1 after compression encoding of these pictures are stored in the host computer 20. Has already been calculated.

  The host computer 20 calculates the target data amount Tj to be allocated to the compressed video data obtained by the encoder 18 compressing and encoding the jth picture of the delayed video data S16 according to the following equation 1, and the calculated target data amount Tj is set in the quantization control circuit 180.

  In Equation 1, Dj is the actual difficulty level data of the jth picture of the video data S12, and R'j is assigned to the jth to (j + L-1) th pictures of the video data S12 and S16. The initial value (R′1) of R′j is the target data amount that can be averaged and assigned to each picture of the compressed video data. Each time the encoder 18 generates one picture of compressed video data, the encoder 18 is updated as shown in Expression 3.

The numerical bit rate in Equation 3 indicates the data amount (bit amount) per second determined based on the transmission capacity of the communication line and the recording capacity of the recording medium. ) Indicates the number of pictures per second (30 pictures / second (NTSC), 25 pictures / second (PAL)) included in the video data, and the numerical value Fj + L is per picture determined according to the picture type. The average amount of data is shown.
The DCT circuit 166 of the encoder 18 performs DCT processing on the j-th picture of the input delayed video data S16 and outputs it to the quantization circuit 168.
The quantization circuit 168 quantizes the frequency domain data of the j-th picture input from the DCT circuit 166 with a quantization value Qj that the quantization control circuit 180 adjusts based on the target data amount Tj, and performs quantization. The data is output to the variable length coding circuit 170.
The variable length coding circuit 170 performs variable length coding on the quantized data of the j-th picture input from the quantization circuit 168, and generates compressed video data VOUT having a data amount substantially close to the target data amount Tj. Output.

Similarly, as shown in FIG. 4B, when the encoder 18 compresses and encodes the (j + 1) -th picture (picture a ′ in FIG. 4C) of the delayed video data S16, The encoder 162 completes the compression encoding of the (j + 1) th to (j + L) th pictures (the range c ′ in FIG. 4C) of the video data S12, and the actual difficulty data Dj + 1 of these pictures. , Dj + 2, Dj + 3, ..., Dj + L
Has already been calculated by the host computer 20.

  The host computer 20 calculates the target data amount Tj + 1 to be allocated to the compressed video data obtained by the encoder 18 compressing and encoding the (j + 1) th picture of the delayed video data S16 according to the equation (1). Is set in the control circuit 180.

The encoder 18 compresses and encodes the (j + 1) th picture based on the scale data amount Tj set by the host computer 20 in the quantization control circuit 180, and a compressed video having a data amount close to the target data amount Tj + 1. Data VOUT is generated and output.
Similarly, the video data compression apparatus 1 sequentially compresses and encodes the kth picture of the delayed video data S16 by changing the quantization value Qk (k = j + 2, j + 3,...) For each picture. , And output as compressed video data VOUT.

As described above, according to the video data compression apparatus 1 shown in the first embodiment, the difficulty level of the pattern of the uncompressed video data VIN is calculated in a short time, and adaptively at a compression rate corresponding to the calculated difficulty level. The uncompressed video data VIN can be compressed and encoded. That is, according to the video data compression apparatus 1 shown in the first embodiment, unlike the two-pass encoding method, the non-compressed video is adaptively based on the difficulty of the pattern of the non-compressed video data VIN almost in real time. The data VIN can be compressed and encoded, and can be applied to applications requiring real-time performance such as live broadcasting.
In addition to the one shown in the first embodiment, the data multiplexing apparatus 1 according to the present invention uses the amount of compressed video data compression-encoded by the encoder 162 as difficulty data as it is, and performs processing of the host computer 20. Various configurations such as simplification can be adopted.

Second Embodiment Hereinafter, a second embodiment of the present invention will be described.
The simple two-pass encoding method shown in the first embodiment compresses and encodes input non-compressed video data only by giving a delay of about 1 GOP (for example, 0.5 seconds), and an appropriate data amount. This is an excellent method that can generate compressed video data.

However, these schemes require two encoders. In general, an encoder that compresses and encodes video data requires large-scale hardware, is very expensive even when integrated, and is large in size. Therefore, the need for two encoders in these methods hinders cost reduction, size reduction, and power saving of a device that realizes these methods. Further, it is desirable that the time delay required for the compression encoding is as short as possible, but the actual difficulty level data Dj and the prediction difficulty level data Dj.
Since the calculation process of 'and the preliminary compression encoding process itself require processing time for several pictures, these processes themselves cause a reduction in time delay.

  The second embodiment has been made to solve such a problem, and uses only one encoder and has an appropriate data amount equivalent to the simple 2-pass encoding method and the predictive simple 2-pass encoding method. An object of the present invention is to provide a video data compression method capable of generating compressed video data and having a shorter time delay required for processing.

FIG. 5 is a diagram showing an outline of the configuration of the video data compression apparatus 2 according to the present invention in the second embodiment.
FIG. 6 is a diagram showing a detailed configuration of the compression encoding unit 24 of the video data compression apparatus 2 shown in FIG.
5 and 6, the same components as those of the video data compression device 1 (FIGS. 1 to 3) described in the first embodiment are the same among the components of the video data compression device 2. It is shown with a reference numeral.

As shown in FIG. 5, the video data compression device 2 includes a compression coding unit 10 of the video data compression device 1 (FIGS. 1 to 3) and a compression coding unit 24 in which the encoder 162 is removed from the compression coding unit 10. The encoder control unit 12 is replaced with the encoder control unit 22, and a buffer memory (buffer) 182 is added.
As shown in FIG. 6, the compression encoding unit 24 includes a video rearrangement circuit 220, a scan conversion / macroblocking circuit 222, and a statistic calculation circuit 224. The other components of the compression encoding unit 24 are as follows: The same configuration as that of the compression encoding unit 10 is adopted.

Similar to the encoder control unit 12, the encoder control unit 22 notifies the host computer 20 of the presence or absence of a picture of the uncompressed video data VIN, and further performs preprocessing for compression coding for each picture of the uncompressed video data VIN. I do.
In the encoder control unit 22, the video rearrangement circuit 220 rearranges the input uncompressed video data in the encoding order.

The scan conversion / macroblocking circuit 222 performs picture / field conversion, and performs 3: 2 pull-down processing when the uncompressed video data VIN is video data of a movie.
The statistic calculation circuit 224 is processed by the video rearrangement circuit 220 and the scan conversion / macroblocking circuit 222, and the statistics such as flatness and intra AC from the picture compressed and encoded into the I picture. Calculate the amount.

With these components, the video data compression apparatus 2 uses the statistical amount (flatness, intra AC) of the uncompressed video data and the prediction error amount (ME residual) of the motion prediction of the degree of difficulty of the pattern of the uncompressed video data VIN. Instead, the target data amount Tj is adaptively calculated in the same manner as the video data compression apparatus 1 (FIGS. 1 and 2), and high-precision feedforward control is performed, so that the uncompressed video data VIN is appropriately It compresses and encodes into compressed video data of a proper data amount.
In the video data compression apparatus 2, the target data amount Tj is determined based on the index data detected in advance by the motion detector 14 and the statistic calculation circuit 224 of the encoder control unit 22. The compression encoding method in the compression device 2 will be referred to as a feed forward rate control (FFRC) method.

  The ME residual is defined as a sum of absolute values or a sum of square values of difference values between a picture to be compressed and video data of a reference picture, and is a picture that becomes a P picture and a B picture after compression by the motion detector 14. It represents the speed of motion of the video and the complexity of the picture, and has a correlation with the degree of difficulty and the amount of data after compression, as with flatness.

Since the I picture is compression-encoded without referring to other pictures, the ME residual cannot be obtained, and flatness and intra AC are used as parameters in place of the ME residual.
Further, flatness is a parameter newly defined as an index representing the spatial flatness of the video in order to realize the video data compression apparatus 2, and indicates the complexity of the video. Correlation with difficulty (degree of difficulty) and data amount after compression.
Intra AC is a parameter newly defined as the sum of variance values of video data for each DCT block in the DCT processing unit in the MPEG system in order to realize the video data compression apparatus 2, and is similar to flatness. In addition, the complexity of the video is indexed, and there is a correlation with the difficulty of the video pattern and the amount of data after compression.

Hereinafter, the ME residual, flatness, and intra AC will be described.
In the simple two-pass encoding method and the predictive simple two-pass encoding method described in the first embodiment, the actual difficulty data Dj indicates the difficulty of the picture pattern, and the target data amount Tj is calculated based on the actual difficulty data Dj. The

  Further, in order to bring the data amount of the compressed video data generated by the encoder 18 close to the value indicated by the target data amount Tj, the quantization value Qj is controlled in the quantization circuit 168 (FIGS. 2 and 6). Accordingly, the quantization process is performed in the quantization circuit 168 of the encoder 18 by using the parameters that can be obtained without compressing and encoding the video data and appropriately indicating the complexity (difficulty) of the picture of the video data as in the case of the actual difficulty data Dj. If previously obtained, the encoder 162 (FIGS. 1 and 2) can be omitted and the object of shortening the processing delay time can be achieved. The ME residual, flatness, and intra AC have a strong correlation with the actual difficulty data Dj, and thus are suitable for achieving such an object.

Relationship between ME Residue and Actual Difficulty Data Dj When generating a P picture and a B picture by performing compression encoding processing with reference to other pictures, the motion detector 14 selects a picture (input picture) to be compressed. ) And a macro block that minimizes the sum of absolute values or sums of square values of difference values between the target macro block and the referenced picture (reference picture), and obtains a motion vector. As described above, the ME residual is defined as a value obtained by summing the absolute sum or the square sum of the difference values of the respective macroblocks that are minimized when obtaining the motion vector.

FIG. 7 is a diagram showing the correlation between the ME residual and the actual difficulty data Dj when the P picture is generated by the video data compression apparatuses 1 and 2.
FIG. 8 is a diagram showing a correlation between the ME residual and the actual difficulty data Dj when the B picture is generated by the video data compression apparatuses 1 and 2.
7 and 8, the actual difficulty data Dj uses the data amount of compressed video data obtained by compression encoding using a fixed quantization value by the encoder 18 (hereinafter, FIG. 10, FIG. 10). 7 and 8 are standard images standardized by CCIR [cheer (cheer leaders), mobile (mobile and calender), tennis (table tennis), diva (diva with noise)] and others. 9 is a graph showing the relationship between the ME residual obtained by actually compressing and coding the image (resort) by the MPEG2 method and the actual difficulty data Dj. In FIGS. 7 and 8, the vertical axis of the graph (difficulty) Indicates the actual difficulty data Dj, and the horizontal axis (me resid) indicates the ME residual.
As can be seen with reference to FIGS. 7 and 8, the ME residual has a very strong correlation with the actual difficulty data Dj. Therefore, instead of the actual difficulty data Dj of a picture that becomes a P picture or a B picture after compression, the ME residual is equal to the target data amount Tj.
Can be used to generate

Relationship between Flatness and Actual Difficulty Data Dj FIG. 9 is a diagram illustrating a flatness calculation method.
As shown in FIG. 9, the flatness first divides each DCT block, which is a unit of DCT processing in the MPEG system, into small blocks of 2 pixels × 2 pixels, and then diagonals within these small blocks. The difference value of the pixel data (pixel value) is calculated, the difference value is compared with a predetermined threshold value, and the total number of small blocks whose difference value is smaller than the threshold value is obtained for each picture. Note that the flatness value decreases as the picture pattern is spatially complex, and increases as the image pattern is flat.

FIG. 10 is a diagram showing the correlation between flatness and actual difficulty data Dj when an I picture is generated by the video data compression apparatuses 1 and 2.
FIG. 10 shows the flatness obtained when the standard image standardized by CCIR and other images are actually compression-encoded by the MPEG2 system and the actual difficulty level data Dj, as in FIGS. FIG. 10 is a graph showing the relationship, and in FIG. 10, the vertical axis (difficulty) of the graph indicates actual difficulty data Dj, and the horizontal axis (flatness) indicates flatness.
As shown in FIG. 10, there is a strong negative correlation between the flatness and the actual difficulty data Dj, and the actual difficulty data Dj can be approximated by a method such as substituting the flatness into a linear function. .

Relationship between Intra AC and Actual Difficulty Data Dj Intra AC is calculated for each DCT block as the sum of absolute values of differences between the pixel values of each pixel in the DCT block and the average value of the pixel values in the DCT block. The That is, the intra AC can be obtained by the following expression 4.

FIG. 11 is a diagram showing the correlation between the intra AC and the actual difficulty data Dj when the I picture is generated by the video data compression apparatuses 1 and 2.
As in FIGS. 7 and 8, FIG. 11 shows the relationship between the intra AC and the actual difficulty data Dj obtained when the standard image standardized by CCIR and other images are actually compression-encoded by the MPEG2 system. FIG. 11 is a graph showing the relationship, and in FIG. 11, the vertical axis (difficulty) of the graph shows actual difficulty data Dj, and the horizontal axis (intra AC) shows intra AC.
As shown in FIG. 11, the intra AC and the actual difficulty data Dj have a strong positive correlation, and the actual difficulty data Dj can be approximated by a method such as substituting the intra AC into a linear function. .

As described so far, the actual difficulty data Dj is obtained from each index data (statistic).
Can be approximated by a linear function or the like. Therefore, the actual difficulty data Dj for each picture type can be calculated as shown below.

  The actual difficulty data Dj is approximated by the ME residual according to Equation 5 shown below for the P picture and Equation 6 shown below for the B picture. For the I picture, the actual difficulty level data Dj is approximated by flatness and intra AC or any one of them by an approximate expression similar to Expressions 5 and 6.

  Furthermore, in the simple two-pass encoding method shown in the first embodiment, the target data amount Tj is calculated by substituting the actual difficulty data Dj obtained by these approximations into Equation 1.

Hereinafter, the operation of the video data compression apparatus 2 will be described by taking as an example the case where the actual difficulty data Dj is approximated by ME residual, flatness and intra AC, and the uncompressed video data is compression-encoded by the simple two-pass encoding method. .
In the encoder control unit 22, the video rearrangement circuit 220 rearranges the pictures in the encoding order of the uncompressed video data VIN, the scan conversion / macroblocking circuit 222 performs picture / field conversion and the like, and the statistic calculation circuit 224. Performs calculation processing shown in FIG. 9 and Expression 4 on a picture that is compression-encoded into an I picture, and calculates statistics such as flatness and intra AC.

The motion detector 14 generates a motion vector for a picture that is compression-encoded into a P picture and a B picture, and further calculates an ME residual.
The FIFO memory 160 delays the input video data by L pictures.

The host computer 20 approximates the actual difficulty data Dj by performing the arithmetic processing shown in Expression 5 and Expression 6 on the ME residual generated by the motion detector 14, and performs the same arithmetic processing as Expression 5 and Expression 6. The actual difficulty data Dj is approximated by flatness and intra AC.
Further, the host computer 20 substitutes the approximate actual difficulty data Dj into Equation 1, calculates the target data amount Tj, and sets the calculated target data amount Tj in the quantization control circuit 180 of the encoder 18.

The DCT circuit 166 of the encoder 18 performs DCT processing on the jth picture of the delayed video data.
The quantization circuit 168 quantizes the frequency domain data of the j-th picture input from the DCT circuit 166 with a quantization value Qj that the quantization control circuit 180 adjusts based on the target data amount Tj.
The variable length coding circuit 170 performs variable length coding on the quantized data of the j-th picture input from the quantization circuit 168, and generates compressed video data VOUT having a data amount substantially close to the target data amount Tj. And output to the outside via the buffer memory 182.

  In the TM5 method or the like, a statistic called activity shown in the following equation 7 is used to calculate the quantization value (MQUANT) of the macroblock. Since the activity has a strong correlation with the actual difficulty data Dj as in flatness and intra AC, the activity is used in place of these parameters to approximate the actual difficulty data Dj and perform compression encoding. The data compression device 2 may be configured.

In the above, the operation of the video data compression apparatus 2 has been described by taking the simple two-pass encoding shown in the first embodiment as an example. However, the video data compression apparatus 2 can perform the predictive simple two-pass encoding. Needless to say.
Further, the video data compression apparatus 2 shown in the second embodiment can be modified in the same manner as the video data compression apparatus 1 shown in the first embodiment.

Third Embodiment Prior to the description of the third embodiment of the present invention, the background and purpose of the video data compression apparatus according to the present invention in the third embodiment will be described with reference to FIG.
FIG. 12 shows that the video data compression apparatuses 1 and 2 (FIGS. 1 to 3, 5, and 6) use the compression algorithm shown in TM5 by the MPEG MP @ ML method, It is a figure which shows the evaluation result of the time-dependent change of the occupation amount Bn of a VBV buffer at the time of performing fixed length encoding, keeping data amount (generated bit amount) substantially constant. In FIG. 12, the vertical axis indicates the amount of compressed video data buffered in the VBV buffer, and the horizontal axis indicates the passage of time.

  The compression algorithm shown in TM5 is excellent in that the amount of data per GOP of compressed video data can be made substantially constant. However, in the MPEG fixed rate encoding method in which the data rate of the compressed video data is a fixed value, it is not always necessary to make the data amount constant for each GOP.

  This fixed-rate encoding method satisfies the constraint conditions required by a virtual VBV buffer (video buffering verifier buffer) that buffers video data after compression encoding, that is, the compression buffered in the VBV buffer. The compressed video data is requested only that the amount of video data (occupied amount Bn) does not exceed a prescribed value (causes overflow), and conversely does not fall below the prescribed value (causes underflow).

  When compression encoding is performed using the compression algorithm shown in TM5 by the MP @ ML method, the occupied amount Bn of the compressed video data in the VBV buffer having a buffering capacity of 1.8 Mbit is evaluated. For example, as shown in FIG. In addition, the occupation amount Bn changes at a high value, and it can be seen that the VBV buffer cannot always be used effectively.

The reason why the VBV buffer cannot be used effectively is that the occupied amount Bn in the VBV buffer
The reason for the high value is that although the buffering capacity of the VBV buffer is as large as about 1.8 Mbit, the amount of compressed video data pictures serving as a unit of input / output of the VBV buffer is small.
As described above, when generating compressed video data with a low data rate, the complexity of the data amount of a predetermined number of pictures (GOP) is almost constant regardless of the complexity of the video of the uncompressed video data. The quality of the video obtained by decompressing and decoding the compressed video data obtained by compressing and encoding the uncompressed video data of the simple design part is extremely deteriorated. Conversely, the compressed video data obtained from the simple design part Quality is relatively good. Accordingly, when viewed as a whole, a lot of unevenness occurs in the compressed video data, and the picture becomes unstable and the quality deteriorates.

  The feedback rate control method shown in the third embodiment has been made in view of such problems, and effectively uses the buffering capacity of the VBV buffer within the range of the constraints required by the VBV buffer, and is not compressed. It is an object to improve the quality of compressed video data as a whole by assigning a data amount corresponding to a design to each video data portion.

FIG. 13 is a diagram showing the configuration of the encoder 26 according to the present invention in the third embodiment.
In FIG. 13, the same components as those of the encoder 18 shown in FIGS. 1 to 3, FIG. 5, and FIG.

  The encoder 26 is an apparatus used in place of the encoder 18 of the video data compression apparatus 2 (FIGS. 5 and 6). As shown in FIG. 13, the encoder 26 is a global instead of the quantization control circuit 180. The VBV buffer has a quantization control unit 260 including a complexity calculation circuit (GC calculation circuit) 262, a target data amount calculation (Tj calculation) circuit 264, and a quantization index generation circuit 266. The target data amount Tj and the quantized value Qj (quantized index QIND) can be calculated based on the occupied amount Bn of the compressed video data and the actual difficulty data Dj or the global complexity XI, Xp, XB.

  The encoder 26 performs feedback control with respect to the quantization processing of the quantization circuit 168 based on the data amount of the compressed video data by only one encoder, and the data amount corresponding to the pattern for each portion of the uncompressed video data. Is assigned to generate compressed video data to improve the quality of the compressed video data.

Operation of Each Component Part of Encoder 26 Hereinafter, a part (quantization control unit) of each component part of encoder 26 that is different from encoder 18 of video data compression apparatuses 1 and 2 (FIGS. 1 to 3, 5 and 6). 260) will be described.
Operation of the GC calculation circuit 262 The GC calculation circuit 262 is an average of the data amounts SI, Sp, and SB of the compressed video data output from the variable length coding circuit 170 and the quantization value used by the quantization circuit 168 for quantization. Based on the values QI, Qp, and QB, global complexity XI, Xp, and XB of each picture type is calculated, a target data amount calculation circuit 264, a quantization index generation circuit 266, and, if necessary, the host computer 20 Output for.

The global complexity XI, Xp, XB is calculated for each picture type in the first stage (step 1) of the MPEG TM5 system, and [X (I, P, B); XI = SI QI, Xp = Sp Qp, XB = SB QB, and the global complexity XI, Xp, XB is almost equivalent to the actual difficulty data DI, Dp, DB of I picture, P picture, and B picture, respectively (XI
, Xp, XB≈DI, Dp, DB).

Operation of Target Data Amount Calculation Circuit 264 Overview of Operation (Processing) The target data amount calculation circuit 264 approximates the real difficulty data Dj of each of the global complexity XI, Xp, XB input from the GC calculation circuit 262, and Further, the target data amount Tj of each picture type picture is calculated based on the occupation amount Bn of the VBV buffer to perform rate control. The target data amount Tj calculated by the target data amount calculation circuit 264 is output to the quantization index generation circuit 266.

Calculation Method of Target Data Amount Tj First, a basic calculation method of the target data amount Tj in the target data amount calculation circuit 264 will be described.
As described above, the actual difficulty data Dj of each picture type is almost the same as the global complexity XI, Xp, XB. Therefore, the target data amount calculation circuit 264 can calculate the target data amount Tj for each picture type from the global complexity XI, Xp, XB.
In each of the above relational expressions, the weighting coefficients Kp and KB are coefficients introduced to perform different weighting on the target data amount Tj for each picture type. If the values of the weighting coefficients Kp and KB are increased, respectively. The smaller the target data amount Tj for the I picture, the smaller the target data amount Tj for the P picture and B picture. For example, in the MPEG TM5 system, the weighting coefficients Kp and KB are fixed values, which are 1.0 and 1.4 (Kp = 1.0, KB = 1.4, default values), respectively.

  Thus, in the MPEG TM5 system, the P picture is given a target data amount Tj according to the ratio of the global complexity Xp of the P picture to the global complexity XI of the I picture, A target data amount Tj that is intentionally smaller than the ratio of the global complexity XB of the B picture to the global complexity XI of the I picture is given.

Rate Control Method Next, a rate control method in the target data amount calculation circuit 264 will be described.
A parameter R is a parameter that plays an important role in the rate control of the MPEG TM5 system. This parameter R indicates the amount of data that can be allocated to the remaining pictures in a control unit (eg, GOP) for rate control in the MPEG system.

  Here, in the video data compression apparatuses 1 and 2 (FIGS. 1 to 3, FIG. 5 and FIG. 6), for example, the video of the first half of the GOP is complicated (actual difficulty data Dj and global complexity XI, If a large amount of data is allocated to the first half of the GOP when the values of Xp, XB, etc. are large), the parameter R for the second half of the GOP becomes an extremely small value or becomes a negative number. In some cases, the amount of data to be allocated to the second half of the GOP is insufficient.

  As described above, in the video data compression apparatuses 1 and 2, the value of the parameter R may become extremely small or become a negative number when the host computer 20 (FIGS. 1 and 5) controls the rate control. In order to keep the data amount of each unit GOP constant, the data amount that is excessively allocated to the first half picture of the GOP is compensated by assigning a small data amount to the second half picture of the GOP. This is because the amount of data is allocated. In the host computer 20, the parameter R is thus used for data amount compensation processing in a relatively short period such as GOP.

  On the other hand, the target data amount calculation circuit 264 of the encoder 26 does not perform rate control only with the parameter R for making the data amount constant in such a short control unit, but within the range of the constraint condition of the VBV buffer. The parameter Rj ′ for evenly allocating the remaining data amount is controlled so that the data amount in the long term becomes constant.

That is, the target data amount calculation circuit 264 controls the parameter Rj ′, and the quality of the compressed video data deteriorates when an excessively allocated data amount is allocated to a picture included in a certain period of uncompressed video data and a small data amount is allocated. The target data amount Tj is adjusted so that compensation is made in a period in which the quality of the compressed video data is small even if a small amount of data is allocated, without compensation in such a period.
Furthermore, the target data amount calculation circuit 264 updates the value of the parameter Rj ′ by performing the same processing as in Equation 3 every time the encoder 26 compresses and encodes one picture.

However, when the parameter Rj ′ is adjusted so that the amount of compressed video data is large (the data rate is high), it is difficult to predict the amount of increase in the amount of compressed video data. Underflow can occur. Therefore, when performing rate control so as to increase the data amount of the compressed video data, the target data amount calculation circuit 264 takes into account the VBV buffer occupation amount Bn (compression) in consideration of the future VBV buffer occupation amount Bn. The parameter Rj ′ is adjusted only when the remaining amount of video data is large.

Note that the target data amount calculation circuit 264 further performs the processing described below in order to realize rate control in consideration of the VBV buffer occupation amount Bn described above.
In other words, the target data amount calculation circuit 264 assigns the amount of data allocated to the complex portion of the video data to the data amount until the VBV buffer underflows, not the data rate of the compressed video data output by the encoder 26. Ask based.

  In addition, the target data amount calculation circuit 264 stores the total value (debt amount) of the data amount allocated to the complicated portion of the video data as the parameter sum-supplement (initial value 0). When the sum of the values of the actual difficulty data Dj of the predetermined number of pictures becomes small, the rate control is performed so that the value of the parameter sum-supplement is decreased, and the parameters at the time when the compression encoding of the uncompressed video data is completed. Rate control is performed so that the sum-supplement value becomes a negative value very close to zero. However, when the VBV buffer occupancy Bn is small, the target data amount calculation circuit 264 performs the rate so that the value of the target data amount Tj of each picture of the video data becomes small regardless of the value of the actual difficulty data Dj. Control and prevent underflow.

Summary of processing contents of target data amount calculation circuit 264 Hereinafter, the target data amount Tj by the target data amount calculation circuit 264 will be described in detail with reference to FIG. 14 and mathematical expressions.
FIG. 14 is a flowchart showing processing of the target data amount calculation circuit 264 shown in FIG.
As shown in FIG. 14, in step 500 (S500), the target data amount calculation circuit 264 checks the occupation amount Bn of the VBV buffer, and a sufficient amount of compressed video data is buffered in the VBV buffer. It is determined whether there is room for no flow. If there is room, the process proceeds to S502. If there is no room, the process proceeds to S512.

  Note that the threshold VBV-R'j-Margin shown in Equation 8 below is used to determine the occupation amount Bn of the VBV buffer.

  In Equation 8, last-I-genbit is the data amount of the latest I picture, and VBV-Margin is a constant for countermeasures against underflow in calculating the target data amount Tj. Is the amount of data per picture. As shown in Equation 8, by using the latest I-picture data amount last-I-genbit for calculation of the threshold value VBV-R'j-Margin, the encoder 26 then compresses the I-picture with the largest data amount. Even when video data is generated, the occurrence of underflow can be almost completely prevented. The target data amount calculation circuit 264 compares the occupation amount Bn of the VBV buffer with the threshold value VBV-R'j-Margin to determine whether the VBV buffer has a margin in the processing of S500.

  Further, the target data amount calculation circuit 264 does not necessarily need to determine the VBV buffer occupation amount Bn in the processing of S500 every time the encoder 26 compresses and encodes a picture. For example, the encoder 26 generates a P picture. It may be done only immediately after.

  This is due to the following reason. That is, the occupancy of the VBV buffer decreases immediately after the encoder 26 generates the I picture, but the occupancy usually recovers until the next generation of the I picture. Therefore, immediately after the encoder 26 generates the I picture. Therefore, the target data amount calculation circuit 264 does not need to make a determination in the processing of S500. Conversely, immediately after the encoder 26 generates a B picture with a small amount of data, the target data amount calculation circuit 264 performs the processing in the processing of S500. This is because, when the determination is made, it is erroneously determined that there is enough room for the VBV buffer to underflow, and there is a possibility that the VBV buffer will underflow.

  In step 502 (S502), the target data amount calculation circuit 264 determines whether or not the total sum of global complexity XI, Xp, XB of the N pictures shown in the following equation 9-1 is larger than the threshold value Th1. To do. If the sum sum-difficulty value is greater than the threshold value Th1, the process proceeds to S504, and if it is equal to or less than the threshold value Th1, the process proceeds to S508. The threshold value Th1 determines whether to increase the data amount of the compressed video data by increasing the value of the parameter Rj ′, or conversely, to decrease the data amount of the compressed video data by decreasing the value of the parameter Rj ′. Is important to.

  In step 504 (S504), the target data amount calculation circuit 264 determines whether or not the parameter Rj 'is larger than the threshold value (G + Th2) as shown in the following equation 9-2. When the value of the parameter Rj ′ is larger than the threshold (G + Th2), the process proceeds to S506, and when it is smaller than the threshold (G + Th2), the process proceeds to S516 (G = N × bit-rate / picture-rate). .

  In step 506 (S506), the target data amount calculation circuit 264 calculates a data amount (supplemented data amount) supplement to be added (supplemented) to the parameter Rj 'by the following equation 10-1. The parameter β (0 <β <1) in Expression 10-1 is defined as shown in Expression 10-2, and is a parameter for determining the amount of data until the VBV buffer underflows. The larger the value of the parameter β and the larger the margin for underflow of the VBV buffer, the larger the value of the supplemental data amount supplement.

  In addition, the threshold value Th3 in Expression 10-1 is a constant for determining the value of the supplemental data amount supplement, and MAX-supplement is a limit value for limiting the supplemental data amount supplement.

  When the sum sum-difficulty value becomes larger than (Th1 + Th3), the value of the fractional term on the right side of Equation 10-1 becomes larger than 1, so the value of the supplemental data amount supplement as shown in Equation 11 below. Correct.

  In step 508 (S508), the target data amount calculation circuit 264 has a parameter sum-supplement having a positive value, and the replenishment data amount supplement that has been replenished to a complex part of the image of the video data is not completely compensated ( Determine if you have debt. If there is a debt, the process proceeds to S510, and if there is no debt, the process proceeds to S512.

  In step 510 (S510), the target data amount calculation circuit 264 sets the value of the parameter β in the equation 10-1 to 1 in order to compensate for the supplement data amount supplement supplied to the complex part of the picture of the video data. The negative supply data amount supplement shown in Equation 12 is calculated. By adding the negative supply data amount supplement to the parameter Rj '(S514), the data amount of the compressed video data is reduced, and the parameter sum-supplement can be brought close to 0 (repayment of debt).

  In step 512 (S512), the target data amount calculation circuit 264 determines that there is a possibility of underflow in the VBV buffer, and calculates a negative supply data amount supplement according to Equation 13 below. By adding the negative supplemental data amount supplement to the parameter Rj '(S514), the data amount of the compressed video data is reduced and underflow of the VBV buffer is prevented.

  In step 514 (S514), the target data amount calculation circuit 264 updates the parameters Rj ′ and sum-supplement according to the following equations 14 and 15.

  In step 516 (S516), the target data amount calculation circuit 264 calculates the target data amount Tj as shown in the following equation 16 and outputs the target data amount Tj to the quantization index generation circuit 266.

  In Equation 16, NI, Np, and NB indicate the number of I-pictures, P-pictures, and B-pictures that appear in one GOP, respectively. When the configuration of 1GOP is N = 1 and M = 3, NI = 1, Np = 4, NB = 10.

  In step 518 (S518), the quantization index generation circuit 266 generates a quantization index QIND based on the target data amount Tj generated by the target data amount calculation circuit 264, and outputs it to the quantization circuit 168.

In step 520 (S520), the components other than the quantization control unit 260 of the encoder 26 compress and encode the uncompressed video data based on the quantization index QIND generated by the quantization index generation circuit 266.
In step 522 (S522), the target data amount calculation circuit 264 increments the variable j.

Operation of Quantization Index Generation Circuit 266 Hereinafter, the operation (processing) of the quantization index generation circuit 266 will be described with reference to FIG. 13 again.
The quantization index generation circuit 266, for example, similarly to the second stage and the third stage (step 2 and step 3) of MPEG TM5, the target data amount Tj input from the target data amount calculation circuit 264, and A quantization index QIND is generated from the global complexity XI, Xp, XB input from the GC calculation circuit 262 and output to the quantization circuit 168.

  The quantization index is data used as an index indicating a combination of quantization values Qj that change for each macroblock that is a unit of quantization processing in the quantization circuit 168, and is equivalent to the quantization value Qj. is there. That is, the quantization circuit 168 that receives the quantization index from the quantization index generation circuit 266 converts the quantization index Qj indicated by the received quantization index into a combination, and quantizes the video data input from the DCT circuit 166. To do.

Hereinafter, the operation of the encoder 26 (FIG. 13) will be described.
The motion detector 14 performs processing such as generation of a motion vector, as in the first embodiment.
The encoder control unit 22 performs a picture rearrangement process and the like as in the first embodiment.

Each time the encoder 26 (FIG. 13) finishes compression encoding for one picture, the GC calculation circuit 262 of the quantization control unit 260 calculates the average of the quantization values Qj from the quantization index of the quantization index generation circuit 266. A value is calculated, and global complexity XI, Xp, XB is calculated from the average value of the quantized value Qj and the data amount Sj of the compressed video data.
The target data amount calculation circuit 264 calculates the target data amount Tj of each picture type generated most recently, as described with reference to FIG. 14.

The quantization index generation circuit 266 calculates a quantization index based on the calculated target data amount Tj and global complexity XI, Xp, XB, and sets the quantization index in the quantization circuit 168 of the encoder 26.
The DCT circuit 166 performs DCT processing on the next picture in the same manner as in the first and second embodiments.

The quantization circuit 168 converts the set quantization index into a quantized value Qj for the DCT-processed video data, and performs a quantization process using the obtained quantized value Qj.
The variable-length encoding circuit 170 performs variable-length encoding in the same manner as in the first and second embodiments, generates compressed video data having a data amount that is substantially close to the target data amount Tj, and a buffer. The data is output via the memory 182.

The contents of the processing of the encoder 26 shown as the third embodiment are the same as the video data compression apparatuses 1 and 2 shown in the first embodiment and the second embodiment (FIGS. 1 to 3, FIG. 5, FIG. It is also applicable to 6).
Further, even if the target data amount calculation circuit 264 of the encoder 26 is configured to calculate the target data amount Tj using the actual difficulty data Dj, the target data amount Tj is calculated using the global complexity XI, Xp, XB. It may be calculated.

In addition, the processing performed by the quantization control unit 260 in the encoder 26 can be performed by the host computer 20 in the video data compression apparatuses 1 and 2 (FIGS. 1 to 3, 5 and 6).
The formulas defining the parameters shown in the third embodiment are examples, and the formulas can be changed according to the configuration and application of the encoder 26.
Further, the encoder 26 shown in the third embodiment can be modified as shown in the first embodiment and the second embodiment.

  FIG. 15 shows the occupancy amount Bn of the VBV buffer when the encoder 26 (FIG. 13) performs fixed-length encoding with the GOP data amount of the compressed video data kept substantially constant according to the MPEG MP @ ML method. It is a figure which shows the evaluation result of a time-dependent change. In FIG. 15, the vertical axis indicates the amount of compressed video data buffered in the VBV buffer, and the horizontal axis indicates time.

When the encoder 26 described above performs fixed-length encoding while keeping the GOP data amount of the compressed video data substantially constant, the occupation amount Bn of the VBV buffer occupation amount Bn is obtained.
15 changes in a large range as shown in FIG. 15, compared with the case where the video data compression apparatuses 1 and 2 (FIGS. 1 to 3, 3, 5 and 6) shown in FIG. 12 generate compressed video data. It can be seen that the VBV buffer is effectively used within the range of the constraints required by the VBV buffer.
Also, according to the encoder 26, the quality of the compressed video data can be improved as a whole by allocating a data amount corresponding to the pattern for each portion of the uncompressed video data.

Fourth Embodiment Hereinafter, a feedforward rate control system will be described as a fourth embodiment of the present invention. The feed-forward rate control method effectively uses the buffering capacity of the VBV buffer within the range of constraints required by the VBV buffer, and assigns a data amount corresponding to the design for each portion of the uncompressed video data, The object is to improve the quality of compressed video data as a whole.

FIG. 16 is a diagram showing the configuration of the video data compression apparatus 4 according to the present invention in the fourth embodiment.
FIG. 17 is a diagram showing a configuration of the encoder 28 shown in FIG.
18 is a diagram illustrating a configuration of the quantization control unit 280 illustrated in FIG.
16 to 18, the configuration of the video data compression apparatuses 1 and 2 and the encoder 26 shown in FIGS. 1 to 3, 5, 6, and 13 among the components of the video data compression apparatus 4. The same reference numerals are given to the same parts.

As shown in FIG. 16, the video data compression apparatus 4 employs a configuration in which the encoder 18 of the video data compression apparatuses 2 and 3 (FIGS. 5, 6, and 13) is replaced with an encoder 28.
Also, as shown in FIG. 17, the encoder 28 employs a configuration in which the quantization control circuit 180 is replaced with a quantization control unit 280. As shown in FIG. 18, the quantization control unit 280 receives the actual difficulty data (Dj ) A calculation circuit 282, a target data amount (Tj) calculation circuit 284, a parameter (Rj ') calculation circuit 286, and a quantization index generation circuit 288.

  As in the encoder 26 (FIG. 13), the quantization control unit 280 does not depend on the host computer 20, and the index data [statistics; flatness described in the second embodiment (FIGS. 9 and 10), Intra AC (FIG. 11), activity (formula 7) and ME residual (FIGS. 7 and 8)], and the compressed video data occupation amount Bn in the VBV buffer, the target data amount Tj and the quantized value Qj ( The quantization index QIND) can be calculated.

  The video data compression apparatus 4 performs feedforward control for the quantization processing of the quantization circuit 168 based on the data amount of the compressed video data by only one encoder by using these components, and the image data is compressed for each portion of the uncompressed video data. Corresponding data amount is allocated to generate compressed video data to improve the quality of the compressed video data.

Operation of Each Component of Video Data Compressor 4 In the following, video data compressor 1, 2, 3 (FIGS. 1 to 3, 5, 5, 6 and 13) of each component of video data compressor 4 The operation of the different part (quantization control unit 280) will be described.
Actual difficulty level data calculation circuit 282 calculation circuit The actual difficulty level data calculation circuit 282 is an index data input from the motion detector 14 (by approximation by ME residual, as shown in Equations 5 and 6, P picture and B picture The actual difficulty level data Dj is calculated and approximated by the index data (flatness, intra AC, and activity) input from the statistic calculation circuit 224 of the encoder control unit 22 to obtain the I picture as in Expressions 5 and 6. The actual difficulty level data Dj is calculated and output to the parameter calculation circuit 286 and the parameter calculation circuit 286.

Operation of Target Data Amount Calculation Circuit 284 The target data amount calculation circuit 284 performs the processing shown in Expression 1 in the first embodiment, as with the target data amount calculation circuit 264 of the encoder 26 (FIG. 13), and the actual difficulty level. Based on the actual difficulty data Dj input from the data calculation circuit 282 and the parameter Rj ′ input from the parameter calculation circuit 286, the target data amount Tj of each picture of each picture type is calculated to perform rate control.

Operation of Parameter Calculation Circuit 286 The parameter calculation circuit 286 adjusts the parameter Rj ′ by performing the processing shown in Expressions 8 to 15 and FIG. 14 in the same manner as the target data amount calculation circuit 264 (FIG. 13) of the encoder 26. ,Update. However, the parameter calculation circuit 286 calculates the target data amount Tj according to Equation 1 instead of Equation 16 in the processing of S516 shown in FIG. 14 and outputs it to the quantization index generation circuit 288.

Operation of Quantization Index Generation Circuit 288 The quantization index generation circuit 288 is based on the target data amount Tj input from the target data amount calculation circuit 284, similarly to the quantization index generation circuit 266 (FIG. 13) of the encoder 26. A quantization index QIND is generated and output to the quantization circuit 168.

Hereinafter, the operation of the video data compression apparatus 4 will be described.
The actual difficulty level data calculation circuit 282 of the quantization control unit 280 uses the index data (ME residual, flatness, intra AC, and activity) input from the motion detector 14 and the encoder control unit 22 to formulas 5 and 6. As shown, the actual difficulty data Dj is calculated.

As shown in Expressions 8 to 15, the parameter calculation circuit 286 adjusts the parameter Rj ′ in accordance with the occupation amount of the VBV buffer and the complexity of the picture of the video data, and performs rate control.
The target data amount calculation circuit 284 calculates the target data amount Tj by substituting the parameter Rj ′ adjusted by the parameter calculation circuit 286 into Equation 1.

The quantization index generation circuit 288 calculates a quantization index QIND from the calculated target data amount Tj.
The parts other than the quantization control unit 280 of the encoder 28 compress and encode the uncompressed video data using the quantization index QIND calculated by the parameter calculation circuit 286.

The contents of the processing of the video data compression device 4 shown as the fourth embodiment are the same as those of the video data compression devices 1 and 2 shown in the first to third embodiments (FIGS. 1 to 3 and FIG. 3). 5 and FIG. 6).
Further, the processing performed by the quantization control unit 280 in the video data compression apparatus 4 can be performed by the host computer 20 in the video data compression apparatuses 1 and 2 (FIGS. 1 to 3, 5 and 6). .
Also, the video data compression apparatus 4 shown in the fourth embodiment can be modified as shown in the first to third embodiments.

Fifth Embodiment Hereinafter, a modified example of the operation of the encoder 26 shown in the third embodiment will be described as a fifth embodiment of the present invention.
Up to this point, the simple two-pass encoding method is described in the first embodiment, and the FFRC method is described in the second embodiment. Furthermore, in the third and fourth embodiments, the VBV buffer occupancy is increased. A feedback rate control method and a feed forward rate control method for adjusting the data amount of compressed video data have been described.

  The TM 5 of the MPEG method calculates the target data amount Tj using the parameter R, and each method shown in the first to fourth embodiments uses the parameter Rj ′ (Equation 1 or the like). When each of these methods compresses and encodes a portion of uncompressed video data that has a very difficult picture pattern (high coding difficulty) into compressed video data having a low data rate, the quantization value Qj ( Even if an attempt is made to increase the compression ratio by increasing the value of the quantization index QIND) and reduce the data amount, the data amount of the actually generated compressed video data exceeds the target data amount Tj, and the values of the parameters R and Rj ′ are The value decreases rapidly, and the values of the parameters R and Rj ′ may become 0 or less in the last picture of the rate control unit (for example, GOP).

  For example, in the TM5 of MPEG, when the value of the parameter R becomes 0 or less, each picture has the minimum amount of data (frame-bit / 8; where frame-bit is the data per desired picture of compressed video data. Rate) will be assigned. As described above, when a picture to which the minimum amount of data is allocated is compressed and encoded into compressed video data having a data rate as low as 1/8 of a desired data rate, the quality of the compressed video data obtained from such a portion becomes remarkable. It will decline.

In addition, for example, if the compression encoding process of uncompressed video data, which is difficult to picture, is continued for a long time, the values of the parameters R and Rj ′ become very small, and the picture of uncompressed video data becomes simple. After that, the values of the parameters R and Rj ′ do not recover to a large positive value for a while, and the minimum amount of data is allocated to each GOP until the values of the parameters R and Rj ′ recover. As a result, the distortion of the compressed video data increases.
On the other hand, the parameter Rj ′ is originally an average value of the data amount allocated to the L pictures corresponding to the delay time of the FIFO memory 160, and therefore the value does not deviate greatly from (frame-bit × L). .

  The fifth embodiment of the present invention has been made in view of the above-described problems. The image pattern of uncompressed video data is complicated (the value of the actual difficulty data Dj is large), and the target data amount Tj is set. On the other hand, even if the amount of compressed video data Sj actually generated is large, the quality of the compressed video data can be kept high, and the video changes from a complex picture to a simple picture. The processing contents of the target data amount calculation circuit 264 (FIG. 13) of the quantization control unit 260 of the encoder 26 shown in the third embodiment for the purpose of promptly recovering the value of the parameter Rj ′ Is a change.

In the fifth embodiment, as in the third embodiment, the encoder 26 uses the VBV buffer occupancy Bn and the global complexity XI, Xp.
, XB based on feedback control of the target data amount Tj, and further limiting the parameter Rj ′ to be equal to or lower than a predetermined lower limit value, thereby obtaining the same effect as the rate control in the third embodiment, Prevents significant degradation of compressed video data quality.

Operation of Target Data Amount Calculation Circuit 264 The processing contents are different from those of the video data compression apparatuses 1 and 2 and the encoder 26 (FIGS. 1 to 3, 5, 6, and 13) among the components of the encoder 26. The operation (processing contents) of the target data amount calculation circuit 264 will be described.
As in the third embodiment, the target data amount calculation circuit 264 approximates the global difficulty data Dj of each of the global complexity XI, Xp, and XB input from the GC calculation circuit 262, and further, a VBV buffer. Based on the occupation amount Bn, the target data amount Tj of each picture of each picture type is calculated to perform rate control.

Rate Control Method The target data amount calculation circuit 264 adjusts the parameter Rj ′ in consideration of the VBV buffer occupancy as in the third embodiment, and sets the global complexity XI, Xp, XB to the parameter Rj ′. The target data amount Tj is adjusted by multiplying the multiplier calculated from the above.
However, unlike in the third embodiment, in the fifth embodiment, the target data amount calculation circuit 264 sets a lower limit value Rmin for the parameter Rj ′ and calculates in the same manner as in the third embodiment. When the measured parameter Rj ′ is lower than the lower limit value Rmin [Rj ′ <Rmin], [Rj ′ = Rmin] is set, and the parameter Rj ′ is restricted so as not to be lower than the lower limit value Rmin. As the lower limit value Rmin, for example, [Rmin = frame-bit × L × 3/4] or [Rmin = frame-bit × L × 1/4] is used.

  As shown in Expression 3 in the first embodiment, the data amount of the j-th picture is Sj, the data amount of the j + L-th picture is Sj + L, and the parameter Rj ′ depends on the picture type. Is Fj + L, the value of the next parameter Rj + 1 'is (Rj'-Sj + Fj + L) [Rj + 1' = Rj'-Sj + Fj + L] It becomes. However, the next parameter Rj + 1 '(= Rj'-Sj + Fj + L) may also be less than the lower limit value Rmin [Rj + 1' <Rmin]. In this case, the next parameter Rj + 1 'is limited to the lower limit value Rmin as shown in the following equation 17.

  Further, the target data amount calculation circuit 264 stores, as the parameter sum-supplement, the total value (debt amount) of the data amount that is largely allocated to the portion where the video of the video data is complex, as in the third embodiment. Accordingly, when the value of the parameter Rj ′ is not limited to the lower limit value Rmin as described above, the parameter sum-supplement is updated as shown in Equation 15 to limit the value of the parameter Rj ′ to the lower limit value Rmin. In this case, the parameter sum-supplement is updated by accumulating the supplementary data amount supplement as shown in the following equation (18).

Summary of Processing Content of Target Data Amount Calculation Circuit 264 Hereinafter, the rate control processing by the target data amount calculation circuit 264 in the fifth embodiment will be described in detail with reference to FIG.
FIG. 19 is a flowchart showing the processing of the target data amount calculation circuit 264 in the fifth embodiment.
As illustrated in FIG. 19, the target data amount calculation circuit 264 performs the same processes as the processes illustrated in FIG. 14 in the third embodiment.

  In step 600 (S600), the target data amount calculation circuit 264 proceeds to the processing of S602 or S612 according to the occupation amount Bn of the VBV buffer. Note that the target data amount calculation circuit 264 may determine the VBV buffer occupation amount Bn in the processing of S600 only immediately after the encoder 26 generates the P picture.

In step 602 (S602), the target data amount calculation circuit 264 determines whether or not the sum sum-difficulty value of the actual difficulty data Dj of the N pictures is greater than the threshold value Th1 using Expression 9-1. Accordingly, the process proceeds to S604 or S608.
In step 604 (S604), the target data amount calculation circuit 264 determines whether or not the parameter Rj ′ is larger than the threshold value (G + Th2) using Expression 9-2, and proceeds to the processing of S606 or S616 depending on the determination result. .

In step 606 (S606), the target data amount calculation circuit 264 calculates the supplemental data amount supplement by, for example, Expression 10-1, Expression 10-2, and Expression 11.
In step 608 (S608), the target data amount calculation circuit 264 determines whether or not the replenishment data amount supplement is compensated, and proceeds to the processing of S610 or S612 depending on the determination result.
In step 610 (S610), the target data amount calculation circuit 264 calculates a negative supply data amount supplement using Equation 12 to compensate for the supply data amount supplement.

In step 612 (S612), the target data amount calculation circuit 264 calculates a negative supplementary data amount supplement according to Equation 13, and prevents underflow of the VBV buffer.
In step 614 (S614), the target data amount calculation circuit 264 calculates the parameters Rj ′ and sum-supplement using the equations 14 and 15, and the parameter Rj.
If 'is below the lower limit value Rmin, the parameter Rj' is limited to the lower limit value Rmin.

In step 616 (S616), the target data amount calculation circuit 264 calculates the target data amount Tj as shown in Expression 16.
In step 618 (S618), the encoder 26 performs compression encoding processing using the quantization index QIND.
In step 620 (S620), the target data amount calculation circuit 264 calculates and updates the next parameter Rj + 1 ′ by Expression 3.

In step 622 (S622), the target data amount calculation circuit 264 determines whether or not the next parameter Rj + 1 ′ is larger than the lower limit value Rmin. If the next parameter Rj + 1 'is larger than the lower limit value Rmin, the process proceeds to S628. If not, the process proceeds to S624.
In step 624 (S624), the target data amount calculation circuit 264 limits the next parameter Rj + 1 'to the lower limit value Rmin.

In step 626 (S626), the target data amount calculation circuit 264 updates the parameter sum-supplement by Expression 18.
In step 628 (S628), the target data amount calculation circuit 264 increments the variable j.

Hereinafter, the operation of the encoder 26 (FIG. 13) in the fifth embodiment will be described.
The motion detector 14 performs processing such as generation of a motion vector, as in the first and third embodiments.
The encoder control unit 22 performs a picture rearrangement process and the like as in the first embodiment.
The FIFO memory 160 delays the input video data by L pictures as in the first embodiment.

Each time the encoder 26 (FIG. 13) finishes compression encoding for one picture, the GC calculation circuit 262 of the quantization control unit 260 calculates the average of the quantization values Qj from the quantization index of the quantization index generation circuit 266. A value is calculated, and global complexity XI, Xp, XB is calculated from the average value of the quantized value Qj and the data amount Sj of the compressed video data.
The target data amount calculation circuit 264 is the same as the target data amount calculation circuit 264 for compressed video data, as described with reference to FIG. 19 based on the global complexity XI, Xp, and XB of the most recently generated picture types. Then, the target data amount Tj of the next picture is calculated.

The quantization index generation circuit 266 calculates a quantization index based on the calculated target data amount Tj and global complexity XI, Xp, XB, and sets the quantization index in the quantization circuit 168 of the encoder 26.
The DCT circuit 166 performs DCT processing on the next picture, as in the first embodiment.

The quantization circuit 168 converts the set quantization index into a quantized value Qj for the DCT-processed video data, and performs a quantization process using the obtained quantized value Qj.
The variable length coding circuit 170 performs variable length coding as in the first embodiment, etc., generates compressed video data with a data amount that is substantially close to the target data amount Tj, and passes through the buffer memory 182. Output.

Modified Examples Hereinafter, modified examples of the fifth embodiment will be described.
The improved feedback rate control method shown in the fifth embodiment is the video data compression apparatus 1, 2, 4 (FIGS. 1 to 3) shown in the first embodiment, the second embodiment, and the fourth embodiment. 5, FIG. 6 and FIGS. 16 to 18). In the fifth embodiment, the case where the target data amount calculation circuit 264 calculates the target data amount Tj in consideration of the VBV buffer has been described. However, the target data amount Tj is generated without considering the VBV buffer. Thus, the operation of the target data amount calculation circuit 264 may be changed.

Hereinafter, a modification in which the operation of the video data compression apparatus 1 (FIGS. 1 to 3) is changed and the improved feedback rate control shown in the fifth embodiment is applied will be described with reference to FIG.
FIG. 20 is a flowchart showing processing when the operation of the video data compression apparatus 1 (FIGS. 1 to 3) is changed and the improved feedback rate control shown in the fifth embodiment is performed.
As shown in FIG. 20, the host computer 20 of the video data compression apparatus 1 does not perform rate control in consideration of the VBV buffer, so the processing corresponding to S600 to S614 shown in FIG. 19 is not performed, and S616 to 628 are performed. Only the corresponding process is performed.

In step 700 (S700), the host computer 20 of the video data compression apparatus 1 calculates the target data amount Tj according to Equation 1.
In step 702 (S702), the encoder 18 performs compression encoding processing using the quantization index QIND.
In step 704 (S704), the host computer 20 calculates and updates the next parameter Rj + 1 ′ according to Equation 3.

In step 706 (S706), the host computer 20 determines whether or not the next parameter Rj + 1 'is greater than the lower limit value Rmin, and proceeds to the processing of S712 or S608 depending on the determination result.
In step 708 (S708), the host computer 20 limits the next parameter Rj + 1 'to the lower limit value Rmin.

In step 710 (S710), the host computer 20 updates the parameter sum-supplement according to Equation 18.
In step 712 (S712), the host computer 20 increments the variable j.
In the video data compression apparatus 4 (FIGS. 16 to 18), the feed forward rate control shown in the fourth embodiment is improved and is equivalent to the improved feed forward rate control shown in the fifth embodiment. In order to obtain the effect, the operation of the parameter calculation circuit 286 of the video data compression device 4 may be changed and each process shown in FIG. 14 may be executed. However, in this case, in the process of S616, it is necessary to calculate the target data amount Tj using Equation 1 instead of Equation 16.

In the processing shown in FIG. 20, the improved feedback rate control method can be applied to the MPEG TM5 itself by replacing the parameter Rj ′ with the parameter R in the MPEG TM5.
However, the parameter R in MPEG TM5 takes a large value for the picture of the first part of the GOP, but is almost close to 0 for the end part of the GOP. Although it is possible to set a fixed negative lower limit value Rmin [for example, Rmin = −2 × frame-bit] to the parameter R having such properties, the effect is weak.

  Therefore, when the improved feedback rate control method is applied to MPEG TM5 itself, as shown in FIG. 21, by introducing a function for determining the lower limit value Rmin, the same effect as in the fifth embodiment can be obtained. Obtainable.

  That is, in MPEG TM5, the value of parameter R approaches 0 for the picture of the end part so that the parameter R becomes larger for the picture of the first part of the GOP. As shown, the lower limit value Rmin is (N / 2 × frame-bit) at the beginning of the GOP, and the lower limit value Rmin is (−N / 2 × frame-bit) at the end of the GOP. When the parameter R falls below this straight line, the parameter R is limited to the lower limit value Rmin on the straight line, and the difference value is stored as another parameter, as in the improved feedback rate control system shown in the fifth embodiment. You just have to.

In addition, the host computer 20 can perform the processing performed by the quantization control unit 260 of the encoder 26 in the fifth embodiment.
The formulas defining the parameters shown in the fifth embodiment are examples, and the formulas can be changed in accordance with the configuration and application of the encoder 26.

As described above, according to the improved feedback rate control system shown in the fifth embodiment, when the picture pattern of the input video data is difficult to the data rate after compression and the amount of data becomes too large. However, it is possible to perform rate control while maintaining the distribution of the data amount according to the picture type, and the quality of the compressed video data can be improved.
In addition, since the lower limit value is set, even when the image pattern of difficult input video data becomes simple, the parameters R and Rj 'are restored so that a large amount of data is allocated to the compressed video data within a short time. And the occurrence of uneven quality of the compressed video data can be prevented.

Sixth Embodiment Hereinafter, as a sixth embodiment of the present invention, a modified example (an improved feedforward rate control system) of the operation of the video data compression apparatus 4 (FIG. 16) shown in the fourth embodiment will be described.
The improved feedforward rate control method is the same as the feedforward rate control method shown in the fourth embodiment in the case where the value of the data amount Sj of the compressed video data actually generated is larger than the target data amount Tj. In addition, the quality of the compressed video data can be kept high, and the parameter Rj ′ is improved so that the value of the parameter Rj ′ can be quickly recovered when the video changes from a complex picture to a simple picture.

  In the sixth embodiment, the video data compression apparatus 4 performs feedforward control on the target data amount Tj based on the VBV buffer occupation amount Bn and the index data (ME residual, flatness, intra AC, and activity), and By restricting the parameter Rj ′ from being equal to or lower than a predetermined lower limit, the same effect as the rate control in the fourth embodiment is obtained, and a significant deterioration in the quality of the compressed video data is prevented.

Operation of Each Component Part Hereinafter, a target data amount calculation circuit 284 and a parameter calculation circuit of the quantization control unit 280 (FIG. 17) whose processing contents are different from those in the video data compression device 4 among the respective component parts of the video data compression device 4 The operation (processing contents) of 286 (FIG. 18) will be described.
Operation of Target Data Amount Calculation Circuit 284 The target data amount calculation circuit 284 includes the actual difficulty data Dj (DI, Dp, DB) calculated from the index data by the actual difficulty data calculation circuit 282, and the parameter calculation circuit 286 occupies the VBV buffer. The target data amount Tj for each picture type is calculated based on the amount Bn and the parameter Rj ′ calculated from the actual difficulty data Dj.

Operation of Parameter Calculation Circuit 286 Rate Control Method The parameter calculation circuit 286 performs rate control by adjusting the value of the parameter Rj ′ in consideration of the occupation amount of the VBV buffer, as in the fourth embodiment.
However, the parameter calculation circuit 286 has a lower limit value Rmin for the parameter Rj ′.
Is set to [Rj ′ = Rmin] when the parameter Rj ′ is lower than the lower limit value Rmin [Rj ′ <Rmin], and the parameter Rj ′ is restricted so as not to be lower than the lower limit value Rmin. As the lower limit value Rmin, for example, [Rmin = frame-bit × L × 3/4] or [Rmin = frame-bit × L × 1/4] is used.

  As shown in Equation 3, the data amount of the jth picture is Sj, the data amount of the j + Lth picture is Sj + L, and the data amount to be added to the parameter Rj ′ according to the picture type Is Fj + L, the value of the next parameter Rj + 1 ′ is [Rj + 1 ′ = Rj′−Sj + Fj + L]. However, the next parameter Rj + 1 '(= Rj'-Sj + Fj + L) may also be less than the lower limit value Rmin [Rj + 1' <Rmin]. In this case, the next parameter Rj + 1 'is limited to the lower limit value Rmin as shown in Expression 17.

  The parameter calculation circuit 286 stores the borrowed amount as a parameter sum-supplement. Accordingly, when the value of the parameter Rj ′ is not limited to the lower limit value Rmin as described above, the parameter sum-supplement is updated as shown in Equation 15 to limit the value of the parameter Rj ′ to the lower limit value Rmin. In this case, the parameter sum-supplement is updated by accumulating the supplementary data amount supplement as shown in Expression 18.

Summary of Processing Contents of Parameter Calculation Circuit 286 Hereinafter, with reference to FIG. 19 again, details of rate control processing and related portion processing by the parameter calculation circuit 286 in the sixth embodiment will be described in detail.
In step 600 (S600), the parameter calculation circuit 286 proceeds to the processing of S602 or S612 according to the occupation amount Bn of the VBV buffer. Note that the parameter calculation circuit 286 may determine the VBV buffer occupation amount Bn in the processing of S600 only immediately after the encoder 28 generates the P picture.

In step 602 (S602), the parameter calculation circuit 286 determines whether or not the sum sum-difficulty value of the actual difficulty data Dj of the N pictures is larger than the threshold value Th1 according to Expression 9-1, and according to the determination result. The process proceeds to S604 or S608.
In step 604 (S604), the parameter calculation circuit 286 determines whether or not the parameter Rj ′ is larger than the threshold value (G + Th2) using the equation 9-2, and proceeds to the processing of S606 or S616 depending on the determination result.

In step 606 (S606), the parameter calculation circuit 286 calculates the supplement data amount supplement by, for example, Expression 10-1, Expression 10-2, and Expression 11.
In step 608 (S608), the parameter calculation circuit 286 determines whether or not the supplement data amount supplement is compensated, and proceeds to the processing of S610 or S612 depending on the determination result.
In step 610 (S610), the parameter calculation circuit 286 calculates a negative supply data amount supplement using Equation 12 to compensate for the supply data amount supplement.

In step 612 (S612), the parameter calculation circuit 286 calculates a negative supplemental data amount supplement using Equation 13 to prevent underflow of the VBV buffer.
In step 614 (S614), the parameter calculation circuit 286 calculates the parameters Rj ′ and sum-supplement using the equations 14 and 15, and if the parameter Rj ′ is less than or equal to the lower limit value Rmin, the parameter Rj ′ is set to the lower limit value. Limited to Rmin.

In step 616 (S616), the target data amount calculation circuit 284 is different from the target data amount calculation circuit 264 of the encoder 26 shown in the fifth embodiment, using the equation 1 instead of the equation 16, and the target data amount Tj. Is calculated.
In step 618 (S618), the encoder 28 performs compression encoding processing using the quantization index QIND.
In step 620 (S620), the parameter calculation circuit 286 calculates and updates the next parameter Rj + 1 ′ using Equation 3.

In step 622 (S622), the parameter calculation circuit 286 determines whether or not the next parameter Rj + 1 'is greater than the lower limit value Rmin. If the next parameter Rj + 1 'is larger than the lower limit value Rmin, the process proceeds to S628. If not, the process proceeds to S624.
In step 624 (S624), the parameter calculation circuit 286 limits the next parameter Rj + 1 'to the lower limit value Rmin.

In step 626 (S626), the parameter calculation circuit 286 updates the parameter sum-supplement according to Equation 18.
In step 628 (S628), the parameter calculation circuit 286 increments the variable j.

The operation of the video data compression apparatus 4 (FIG. 16) in the sixth embodiment will be described below.
The motion detector 14 performs processing such as generation of motion vectors and ME residuals.
The encoder control unit 22 performs processing such as picture rearrangement processing and generation of index data (flatness, intra AC, and activity).
The FIFO memory 160 delays the input video data by L pictures.

Each time the encoder 28 (FIG. 16) finishes compression encoding for one picture, the actual difficulty level data calculation circuit 282 of the quantization control unit 280 calculates actual difficulty level data Dj.
The parameter calculating circuit 286 calculates the parameter Rj ′ as shown in FIG. 19, and the target data amount calculating circuit 284 is the actual difficulty data Dj (DI, Dp, DB) of the most recently generated picture of each picture type. ) To calculate the target data amount Tj according to Equation 1.

The quantization index generation circuit 288 calculates a quantization index based on the calculated target data amount Tj and sets the quantization index in the quantization circuit 168 of the encoder 28.
The DCT circuit 166 performs DCT processing on the next picture, as in the first embodiment.

The quantization circuit 168 converts the set quantization index into a quantized value Qj for the DCT-processed video data, and performs a quantization process using the obtained quantized value Qj.
The variable length coding circuit 170 performs variable length coding as in the first embodiment, etc., generates compressed video data with a data amount that is substantially close to the target data amount Tj, and passes through the buffer memory 182. Output.

Modification Hereinafter, with reference to FIG. 20 again, a modification in which the operation of the video data compression apparatus 1 (FIGS. 1 to 3) is changed and the improved feedforward rate control shown in the sixth embodiment is applied will be described. To do.
Since the host computer 20 of the video data compression apparatus 1 does not perform rate control considering the VBV buffer, it does not perform the processing corresponding to S600 to S614 shown in FIG. 19 and performs only the processing corresponding to S616 to 628.

In step 700 (S700), the host computer 20 of the video data compression apparatus 1 calculates the target data amount Tj according to Equation 1.
In step 702 (S702), the encoder 18 performs compression encoding processing using the quantization index QIND.
In step 704 (S704), the host computer 20 calculates and updates the next parameter Rj + 1 ′ according to Equation 3.

In step 706 (S706), the host computer 20 determines whether or not the next parameter Rj + 1 'is greater than the lower limit value Rmin, and proceeds to the processing of S712 or S608 depending on the determination result.
In step 708 (S708), the host computer 20 limits the next parameter Rj + 1 'to the lower limit value Rmin.

In step 710 (S710), the host computer 20 updates the parameter sum-supplement according to Equation 18.
In step 712 (S712), the host computer 20 increments the variable j.

In the processing shown in FIG. 20, the improved feedforward rate control method can be applied to the MPEG TM5 itself by replacing the parameter Rj ′ with the parameter R in the MPEG TM5.
However, the parameter R in MPEG TM5 takes a large value for the picture of the first part of the GOP, but is almost close to 0 for the end part of the GOP. Although it is possible to set a fixed negative lower limit value Rmin [for example, Rmin = −2 × frame-bit] to the parameter R having such properties, the effect is weak.

  Therefore, when the improved feedforward rate control method is applied to MPEG TM5 itself, as shown in FIG. 21, by introducing a function for determining the lower limit value Rmin, the same as in the sixth embodiment. An effect can be obtained.

That is, in MPEG TM5, the value of parameter R approaches 0 for the picture of the end part so that the parameter R becomes larger for the picture of the first part of the GOP. Therefore, as illustrated in FIG. In addition, a straight line is drawn so that the lower limit value Rmin is (N / 2 × frame-bit) at the beginning of the GOP and the lower limit value Rmin is (−N / 2 × frame-bit) at the end of the GOP. When the parameter R falls below this straight line, the parameter R is limited to the lower limit value Rmin on the straight line, and the difference value is stored as another parameter, as in the improved feedforward rate control system shown in the sixth embodiment. Just keep it.
Further, the formulas defining the parameters shown in the sixth embodiment are examples, and the formulas can be changed in accordance with the configuration and use of the video data compression apparatus 4.

As described above, according to the improved feedforward rate control method shown in the sixth embodiment, the picture pattern of the input video data is difficult for the data rate after compression, and the data amount becomes too large. The rate control can be performed while maintaining the distribution of the data amount according to the picture type, and the quality of the compressed video data can be improved.
In addition, since the lower limit value is set, even when the image pattern of difficult input video data becomes simple, the parameters R and Rj 'are restored so that a large amount of data is allocated to the compressed video data within a short time. And the occurrence of uneven quality of the compressed video data can be prevented.

It is a figure which shows the structure of the video data compression apparatus which concerns on this invention. It is a figure which shows the structure of the encoder of the simple 2 pass process part shown in FIG. It is a figure which shows the structure of the encoder shown in FIG. (A)-(C) are figures which show the operation | movement of the simple 2 pass encoding of the video data compression apparatus in 1st Embodiment. It is a figure which shows the outline | summary of a structure of the video data compression apparatus based on this invention in 2nd Embodiment. It is a figure which shows the detailed structure of the compression encoding part of the video data compression apparatus 2 shown in FIG. It is a figure which shows the correlation of ME residual at the time of producing | generating P picture by the video data compression apparatus (FIGS. 1-3, FIG. 5, FIG. 6) and real difficulty data Dj. It is a figure which shows the correlation of ME residual at the time of producing | generating B picture by the video data compression apparatus (FIGS. 1-3, FIG. 5, FIG. 6) and real difficulty data Dj. It is a figure which shows the calculation method of flatness. It is a figure which shows the correlation between the flatness at the time of producing | generating I picture by the video data compression apparatus (FIGS. 1-3, FIG. 5, FIG. 6) and the actual difficulty data Dj. It is a figure which shows the correlation of intra AC at the time of producing | generating I picture by a video data compression apparatus (FIGS. 1-3, FIG. 5, FIG. 6) and real difficulty data Dj. When the video data compression apparatus (FIGS. 1 to 3, 5, and 6) performs fixed-length encoding while maintaining the amount of GOP generated in compressed video data almost constant by the MPEG MP @ ML method It is a figure which shows the evaluation result of the time-dependent change of occupation amount Bn of the VBV buffer. It is a figure which shows the structure of the encoder shown in FIG. It is a flowchart figure which shows the process of the target data amount calculation circuit shown in FIG. According to the MPEG @ ML method of MPEG, the change in the VBV buffer occupancy Bn over time when the encoder (FIG. 13) performs fixed-length encoding with the GOP data amount of the compressed video data kept substantially constant. It is a figure which shows an evaluation result. It is a figure which shows the structure of the video data compression apparatus based on this invention in 4th Embodiment. It is a figure which shows the structure of the encoder shown in FIG. It is a figure which shows the structure of the quantization control part shown in FIG. It is a flowchart figure which shows the process of the target data amount calculation circuit in 5th Embodiment. It is a flowchart figure which shows the process in the case of changing operation | movement of a video data compression apparatus (FIGS. 1-3) and performing the improved feedback rate control shown in 5th Embodiment. It is a figure which shows the function which determines the lower limit Rmin used when applying the improved feedback rate control system shown in 5th Embodiment to MPEGTM5 itself.

Explanation of symbols

  1, 2, 4 ... Video data compression device, 10, 24 ... Compression encoding unit, 12, 22 ... Encoder control unit, 14 ... Motion detector, 16 ... Simple 2-pass processing unit, 160 ... FIFO memory, 162, 18 , 26, 28 ... encoder, 260, 280 ... quantization controller, 262 ... GC calculation circuit, 282 ... actual difficulty data calculation circuit, 284, 264 ... target data amount calculation circuit, 286 ... parameter calculation circuit, 266, 288 ... Quantization index generating circuit, 164 ... adder circuit, 166 ... DCT circuit, 168 ... quantizer circuit, 170 ... variable length encoding circuit, 172 ... inverse quantization circuit, 174 ... inverse DCT circuit, 176 ... adder circuit, 178 ... Motion compensation circuit, 180 ... quantization control circuit, 182 ... buffer memory, 20 ... host computer

Claims (18)

A video data compression device that compresses uncompressed video data of a moving image to a data rate that satisfies a condition determined based on a VBV buffer that buffers compressed video data (compressed video data),
Difficulty level data calculating means for calculating difficulty level data indicating the complexity of the video of the uncompressed video data;
Based on the data amount (occupied data amount) of the compressed video data buffered in the VBV buffer and the calculated difficulty data, the data amount (allocation) after compression to a predetermined number of pictures of uncompressed video data Data amount allocation means for allocating (data amount),
Target value calculation means for calculating a target value of the data amount after compression of the uncompressed video data based on the calculated difficulty data and the allocated data amount;
A video data compression apparatus comprising: compression means for compressing the uncompressed video data by a predetermined compression method so that the amount of data after compression becomes the calculated target value. The video data compression apparatus according to claim 1, wherein the difficulty level data calculation unit calculates a data amount after compression of the uncompressed video data as the difficulty level data. The data amount allocating means determines that the more complex the video of the uncompressed video data is, the more complex the video of the uncompressed video data is based on the calculated difficulty data when the occupied data amount of the VBV buffer is equal to or greater than a predetermined threshold. Increase the quantity value,
The target value calculation means increases the value of the target value as the video of the uncompressed video data is more complex, and sets the value of the target value as the video of the uncompressed video data is simpler. The video data compression apparatus according to claim 1. The data amount allocating unit calculates a difference value by subtracting a reference value obtained by equally allocating all the data amounts given to the compressed video data to all the pictures from the data amount of the generated compressed video data. 4. The video data compression apparatus according to claim 3, wherein, when the compression of the uncompressed video data is completed, the allocated data amount is calculated so that the sum of the calculated difference values becomes a value close to a negative value of 0. 5. . The target amount calculation unit calculates the target value by multiplying a value obtained by dividing the difficulty data of the latest picture by the sum of the difficulty data of a predetermined number of pictures and the calculated allocation data amount. Video data compression device. The video according to claim 3, wherein the compression means compresses the uncompressed video data into a picture type sequence including a plurality of types of pictures (I picture, P picture and B picture or a combination thereof) in a predetermined order. Data compression device. The difficulty level data calculating means includes, as the difficulty level data, an ME residual of a picture compressed into a P picture or a B picture, and flatness, intra AC data, activity and global complexity of a picture compressed into an I picture, The video data compression apparatus according to claim 6, wherein a combination of these is calculated. The data amount allocating means adds, as the predetermined threshold value of the occupied data amount of the VBV buffer, an addition value corresponding to the data rate of the generated compressed video data or a fixed addition to the latest I picture data amount The video data compression apparatus according to claim 6, wherein a numerical value obtained by adding values is used. 9. The data amount allocation unit determines whether or not the amount of data occupied by the VBV buffer is equal to or greater than a predetermined threshold immediately after the compression unit compresses the uncompressed video data into a P picture. The video data compression apparatus described in 1. A video data compression method for compressing uncompressed video data of a moving image to a data rate that satisfies a condition determined based on a VBV buffer that buffers compressed video data (compressed video data),
Based on the amount of data until the VBV buffer underflows and the complexity of the picture of the uncompressed video data picture, the amount of data (allocated data) after being compressed into a predetermined number of pictures of the uncompressed video data Amount)
Based on the complexity of the picture of the picture of the uncompressed video data and the calculated allocation data amount, a target value of the data amount after compression of the non-compressed video data is calculated for each picture,
A video data compression method in which the uncompressed video data is compressed by a predetermined compression method so that the amount of data after compression becomes the calculated target value. The video data compression method according to claim 10, wherein a data amount after compression of the non-compressed video data is used as data indicating the complexity of a picture of the picture of the non-compressed video data. When the occupied data amount of the VBV buffer is equal to or greater than a predetermined threshold, based on the calculated difficulty data, the more complex the video of the uncompressed video data, the larger the value of the allocated data amount,
11. The target value is increased as the video of the uncompressed video data is more complex, and the target value is decreased as the video of the uncompressed video data is simpler. Video data compression method. The difference value is calculated by subtracting the reference value in which all the data amount given to the compressed video data is evenly distributed to all the pictures from the picture data amount of the generated compressed video data, and the uncompressed video data 13. The video data compression method according to claim 12, wherein when the compression of is completed, the allocated data amount is calculated so that the sum of the calculated difference values becomes a value close to a negative value of zero. The video data compression method according to claim 12, wherein the uncompressed video data is compressed into a picture type sequence including a plurality of types of pictures (I picture, P picture and B picture, or a combination thereof) in a predetermined order. The video data compression method according to claim 13, wherein the target value is calculated by multiplying the calculated amount of assigned data by a value obtained by dividing the difficulty data of the latest picture by the sum of the difficulty data of a predetermined number of pictures. As the difficulty data, an ME residual of a picture compressed into a P picture or a B picture, and flatness, intra AC data, activity and global complexity of a picture compressed into an I picture, or a combination thereof are calculated. Item 14. The video data compression method according to Item 13. As the predetermined threshold value of the occupied data amount of the VBV buffer, a value obtained by adding an addition value corresponding to the data rate of the generated compressed video data or a fixed addition value to the data amount of the latest I picture is used. The video data compression method according to claim 13. The video data compression method according to claim 17, wherein the determination as to whether or not the amount of data occupied by the VBV buffer is equal to or greater than a predetermined threshold is performed immediately after the uncompressed video data is compressed into a P picture.

Images