JPH10174106A - Video data compressor and its method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To allow an encoder side to automatically detect a preceding picture type and to conduct compression coding while a group of picture(GOP) phase is matched in the case of repeating compression coding of video data. SOLUTION: Sections, motion compensation section 240 to discrete cosine transformation(DCT) section 244 apply motion compensation and DC transformation or the like to input video data to generate a DCT coefficient. A back search section 248 detects whether or not a minimum value is in existence in total sum of residues as the result of division by each quantization step of the DCT coefficient and discriminates that a picture with a minimum value is compression-coded for I picture in a preceding compression coding and discriminates a configuration (number N of pictures and interval M of P pictures) of a GOP at the preceding compression coding. A picture type control section 250 controls a picture rearrangment section 200 based on the discriminated GOP configuration to rearrange pictures of the input video data in a sequence by the compression coding section 20 for compression coding in the same GOP phase as the preceding compression coding.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention
When copying video data (dubbing) by connecting video tape recorders (VTR devices) that record video data compressed and encoded according to a system, expand and decode reproduced video data, and output the video data, in tandem, The present invention relates to a video data compression apparatus and a video data compression method for compressing and encoding each picture of video data that has been decompressed and decoded into a picture of the same type as in the previous compression encoding.

[0002]

2. Description of the Related Art In recent years, in the field of video data compression, motion compensation (M
C; motion compensation) processing and discrete cosine transform (DC
T: Discrete cosine transfer), etc., to improve the coding efficiency by combining with redundancy reduction processing by orthogonal transformation.
As a so-called MC-DCT method, MPEG2 (moving
The picture experts group 2) method is widely used.

In the MPEG2 system, one I-picture (intra coded picture) that can normally expand and decode uncompressed video data without using pixel data of another picture is used.
, And a P picture (predictive coded picture) to be decompressed and decoded using the pixel data of the previous picture,
B that performs decompression decoding using pixel data of the preceding and following pictures
Picture (bi-directionally predictive coded pictu
re) is compressed and encoded in GOP (group of picture) units each including a predetermined number of GOPs.

Here, for example, when transmitting video data between television broadcasting stations or when copying (dubbing) video data using a plurality of video tape recorders (VTR devices), In order to decompress and decode the video data that has been compression-encoded by the MPEG2 system and to perform compression-encoding again, it is necessary to connect a tandem connection with a compression coder (encoder) and a decompression decoder (decoder) in series. May occur.

[0005] As described above, when the encoder and the decoder are connected in tandem and the compression encoding and decompression decoding of the video data are repeated, the quality of the video deteriorates. In particular, since high-quality video is required in a business system such as a television broadcasting station facility, it is strictly required to prevent deterioration of video quality due to compression encoding and decompression decoding by tandem-connected encoders and decoders.

[0006] In order to minimize the degradation of video quality that occurs in tandem-connected encoders and decoders, it is necessary to use the same quantization step in compression encoding and decompression decoding. Must be kept with it.

In the case of performing compression encoding without performing motion compensation processing, for example, as disclosed in JP-A-5-284458 and JP-A-6-319112, By using the quantized step or the quantized step having a multiple relation thereof, the property that the sum of the remainder of the DCT coefficients is minimized is used, and the quantizing step indicating the minimum value is converted to the optimal quantizing step. By using an excellent method called back search, which is obtained as follows, it is possible to make the quantization steps equal between the time of compression encoding and the time of decompression decoding, and to prevent deterioration in video quality.

However, the GOP is composed of an I picture and a B picture.
When a two-frame configuration including one picture each is used, or when a GOP has a 15-frame configuration, first of all, rather than adjusting the quantization step, each time, in order to suppress the deterioration of video quality at the time of tandem connection, It is important to compress and encode the same picture into the same type of picture (picture type), that is, to match the phase of the GOP each time the compression and encoding are performed.

If the phase of the GOP is lost, the above-described back search method cannot be used. In addition, the same picture is compressed and coded as an I picture by expanding and decoding a B picture or a P picture. When compression encoding is performed for a different picture type, the amount of video information is greatly lost every time the compression encoding is performed, and the quality of the video is greatly deteriorated.

In order to cope with such a problem, for example, as disclosed in Japanese Patent Laid-Open No. 6-284414, a picture type and decoded video data are multiplexed and output at the time of decompression decoding. A method has been considered in which the encoder performs compression encoding by referring to the multiplexed picture types and adjusting the phases of the GOPs.

However, according to the method disclosed in Japanese Patent Application Laid-Open No. 6-284414, picture type information is multiplexed into portions other than effective pixels of video data. When a switcher) or a digital VTR device of a different system enters between a tandem-connected encoder and decoder, picture type information may be lost due to blanking or the like.

As described above, when the picture type information is lost or the picture type information is replaced with other information or random data, the next encoder replaces the random data or the like with the picture type information. There is a possibility that the image quality may be erroneously detected and encoded with a random picture type.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem. Even if the picture type information is not specially multiplexed with effective pixel data, the picture type at the time of the previous compression encoding can be converted on the encoder side. Automatic detection, G
It is an object of the present invention to provide a video data compression apparatus and a method thereof capable of performing compression encoding in accordance with an OP phase.

[0014]

In order to achieve the above object, a video data compression apparatus according to the present invention provides a video data compression apparatus which expands compressed video data compressed to a combination of an I picture, a P picture and a B picture. And I picture detection means for detecting a picture that has been compression-encoded into an I picture in the compression processing, and, based on an interval between the detected I pictures, each picture of the video data is Picture type determining means for determining which picture type among I picture, P picture and B picture has been compressed, and video data compressing means for compressing each picture of the video data into the same picture type as in the compression processing And

Preferably, the I picture detecting means includes:
Orthogonal transformation means for orthogonally transforming each picture of the video data for each macroblock to generate orthogonal transformation data, and a plurality of quantization steps for the orthogonal transformation data for each macroblock of each of the generated pictures of the video data. Dividing means; dividing the orthogonal transform data for each macroblock of each picture of the video data by the plurality of quantization steps; and calculating the sum of the respective remainders of the result of the plurality of quantization steps, and the sum of the remainders Detecting means for detecting whether or not the picture of the video data has been compression-coded into an I-picture in the compression processing based on whether or not there is a minimum value of

Preferably, the I picture detecting means includes:
Orthogonal transformation means for orthogonally transforming a part of a macroblock of each picture of the video data to generate orthogonal transformation data, and a plurality of orthogonal transformation data of a part of the generated macroblock of each picture of the video data picture; Division means for dividing by each of the quantization steps;
A remainder sum calculating means for calculating a sum of remainders of respective division results of the orthogonal transformation data of a part of each macroblock of the picture of the video data by the plurality of quantization steps; and a minimum value of the sum of the remainders Detecting means for detecting whether or not the picture of the video data has been compression-encoded into an I-picture in the compression processing based on whether or not the picture has been compressed.

Preferably, the I picture detecting means includes:
The image processing apparatus further includes a quantization step selection unit that selects a quantization step that minimizes a minimum value of the sum of the remainders, wherein the video data compression unit uses a quantization step that quantizes the video data using the selected quantization step. It has a conversion means.

Preferably, the I picture detecting means comprises:
Before the video data compression means compresses the video data, a picture compressed and coded into an I picture in the compression processing is detected.

The video data compression apparatus according to the present invention compresses and encodes uncompressed video data into a combination of an I picture, a P picture and a B picture by an MPEG method or the like. When compression encoding is performed again, the video data is compression-encoded into the same type of picture as in the previous compression encoding.

In the video data compression apparatus according to the present invention, the I picture detection means uses the processing used for the back search to select the picture of the video data which has already been compression-encoded in the previous compression-encoding. A picture that has been compression-encoded into an I-picture is detected. That is, first, the I picture detecting means first performs orthogonal transform such as DCT on the macroblock for each macroblock,
Find the coefficient.

Next, the DCT coefficient of the picture that has been compression-coded into an I picture in the previous compression coding is calculated by the same quantization step as that of the previous compression coding, or an integer multiple of the previous compression coding. Only by dividing by the quantization step, utilizing the fact that the sum of the remainder obtained as a result of the division shows the minimum value, the picture in which the minimum value appears in the sum of the remainders of the DCT coefficients is determined by the I compression in the previous compression encoding. It is determined that the picture has been compression-coded.

The picture type determining means calculates, for example, the number of pictures (N) of the GOP at the time of the previous compression encoding based on the interval between pictures determined to have been compressed and encoded into an I picture at the time of the previous compression encoding. The interval (M) between P pictures is derived, and it is determined which type of picture each picture of the video data has been compression-encoded in the previous compression encoding.

The video data compression means rearranges each picture of the video data into an order suitable for compression-encoding the same type of picture as in the previous compression encoding, and performs the same GOP as in the previous compression encoding. Compression encoding is performed by the phase and compression encoding method.

Further, according to the video data compression method of the present invention, of the pictures of video data obtained by expanding compressed video data which has been compressed into a combination of an I picture, a P picture and a B picture,
A picture compression-encoded into a picture is detected, and each picture of the video data is assigned to any of the picture types in the I picture, P picture and B picture in the compression processing based on the detected interval of the I picture. It is determined whether the video data has been compressed, and each picture of the video data is compressed into the same picture type as in the compression processing.

Preferably, each picture of the video data is orthogonally transformed for each macroblock to generate orthogonal transformation data, and the generated orthogonal transformation data for each macroblock of each picture of the video data is divided into a plurality of quantum data. Dividing the orthogonal transform data for each macroblock of each picture of the video data by the plurality of quantization steps, and calculating the sum of the remainders of each of the results of the plurality of quantization steps, and has a minimum value of the sum of the remainders Whether or not the picture of the video data has been compression-encoded into an I-picture in the compression process is detected based on whether or not the video data has been compressed.

Preferably, a part of a macroblock of each picture of the video data is orthogonally transformed to generate orthogonal transformation data, and the orthogonal transformation data of a part of a macroblock of each picture of the generated video data is preferably transformed. Is divided by each of a plurality of quantization steps, the sum of the remainders of the respective division results of the plurality of orthogonal transformation data of the orthogonal transform data of a part of the macroblock of each picture of the video data is calculated, and the remainder of the remainder is calculated. Based on whether or not there is a minimum value of the sum, it is detected whether or not the picture of the video data has been compression-encoded into an I-picture in the compression processing.

Preferably, a quantization step that minimizes the minimum value of the sum of the remainders is selected, and the video data is quantized using the selected quantization step.

Preferably, before the video data is compressed, a picture compression-encoded into an I picture in the compression process is detected.

[0029]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment Hereinafter, a first embodiment of the present invention will be described.

Background of the First Embodiment For example, when dubbing video data using a plurality of VTR devices between television broadcasting stations, MPEG
In order to decompress and decode video data that has been compression-encoded by the two methods, and to perform compression-encoding again, it may be necessary to connect the encoder and the decoder in tandem. As described above, when the compression encoding and the decompression decoding of the video data are repeated, the quality of the video is deteriorated.

In order to minimize the deterioration of the video quality when the video data is repeatedly subjected to the compression encoding and decompression decoding, the quantization steps used in the compression encoding and decompression decoding by the above-described back search and the like are made the same. Further, it is important to match the phase of the GOP every time the compression encoding is performed.

If the phase of the GOP is lost, the above-described back search method cannot be used, and the same picture is compressed and coded into an I picture by expanding and decoding a B picture or a P picture. When compression encoding is performed for a different picture type, the amount of video information is greatly lost every time the compression encoding is performed, and the quality of the video is greatly deteriorated.

The encoder 2 (2) shown in the first embodiment
a, 2b) are intended to solve the above-described problem. The picture type at the time of the previous compression encoding is automatically set on the encoder side without specially multiplexing the picture type information with valid pixel data. It is configured so that it can be detected and compressed and encoded with the GOP phase matched.

Video Data Processing System 1 Hereinafter, a video data processing system 1 in which an encoder and a decoder for video data are connected in tandem will be described.
An encoder according to the present invention is a video data processing system 1
Used in

FIG. 1 is a diagram showing a configuration of a video data processing system 1 using an encoder 2 according to the present invention. As shown in FIG. 1, the video data processing system 1
A recording device such as a VTR device or a hard disk device,
Alternatively, a recording / transmission device such as a communication line or a transmission line (b
rr channel) Encoders (encoder) via 3a-3c
1, encoder2) 2a, 2b and decoder (decoder1, decoder
2) A configuration in which 4a and 4b are connected in series (tandem connection) is adopted.

Even if the configuration of the video data processing system 1 is not clearly adopted, for example, the input uncompressed video data is compression-encoded by the MPEG2 system or the like, recorded on a VTR tape, and decompressed during reproduction. Even in the case where a plurality of VTR devices that output in a complex form are connected and video data is copied, the path through which the video data passes is substantially the same as that of the video data processing system 1. Also, encoder 2
(2a, 2b) is a GO including a plurality of types of pictures.
When the non-compressed video data is compression-encoded in P, the present invention has an advantageous effect.

In the video data processing system 1, the encoder 2a is, for example, an MC-
According to the DCT method, non-compressed video data (input video data) input from the outside is compression-coded into GOP units including a plurality of types of pictures to generate compressed video data (compression-coded bit stream), The data is transmitted to the decoder 4a via the recording / transmission device 3 (3a).

The decoder 4a decompresses and decodes the compressed video data input from the encoder 2a via the recording / transmission device 3a, and performs, for example, uncompression (full bit) for a D-1 digital VTR suitable for editing and the like. ) To the encoder 2b via the recording / transmission device 3 (3b).
To be transmitted. The encoder 2b is connected to the decoder 4a via the recording / transmission device 3b in the same manner as the encoder 2a.
The compression and encoding of the full-bit video data input from the device is performed and transmitted to the decoder 4b via the recording / transmission device 3 (3c). The decoder 4b, like the decoder 4a, decompresses and decodes the compressed video data input from the encoder 2b via the recording / transmission device 3c and outputs it as output video data.

In the video data processing system 1, the tandem-connected encoder 2, recording / transmission device 3, and decoder 4 may be included in three or more stages as necessary. Each component performs compression encoding, transmission, and decompression decoding of video data in the same manner as before the second stage.

In FIG. 1, the recording / transmission device 3 described as a full bit channel (full bit channel)
As b, in addition to a device that simply transmits or records / reproduces full-length video data that has been expanded and decoded, for example, digital-to-analog (D / A) conversion of expanded and decoded video data and records it in an analog VTR device Then, a device for converting analog / digital (A / D) to digital video data again, or a device for editing and processing video data via a switcher or a special effect device is also included. Even in such an apparatus, the picture type information recorded together with the video data is lost, so that the encoder 2 (2a, 2b) has a characteristic effect of the present invention.

The encoder 2 (2a, 2b) FIG. 2 is a diagram showing the configuration of an encoder 2 according to the present invention in the first embodiment shown in FIG. 1 (2a, 2b). As shown in FIG. 2, the encoder 2 includes a compression encoding unit 20 and a compression control unit 24.

The compression encoding unit 20 includes a picture rearranging unit 200, a scan conversion blocking unit 202, and a motion detecting unit 2.
04, FIFOs 206 and 220, subtraction circuit 207, DC
T section 208, quantization section 210, variable length coding section (VL
C) 212, inverse quantization unit 214, inverse DCT (IDCT)
216, an adder 218, and a motion compensator 222. The compression control unit 24 includes a motion compensation unit 240, a subtraction circuit 242, a DCT unit 244, a prediction unit 246, a back search unit 248, and a picture type control unit 250.

With these components, the encoder 2 detects the GOP phase in the previous compression encoding from input video data (video in) that has already undergone compression encoding and decompression decoding by the MPEG2 system or the like at least once. , The quantization step in the previous compression encoding is reproduced by the back search method, and the input video data is compression-encoded again with the same quantization step and GOP phase as the previous one,
An encoder and a decoder are connected in tandem to prevent degradation of video quality when compression encoding and decompression decoding of video data are repeated.

Each component of encoder 2 will be described below.

Compression encoding unit 20 The compression encoding unit 20 performs motion compensation processing on input video data, as in a general MPEG2 encoder.
DCT processing and variable length coding processing are performed to perform compression coding. In the compression encoding unit 20, the picture rearranging unit 200, under the control of the picture type control unit 250, selects an order suitable for compression encoding according to the picture type of the picture of the input video data after compression encoding. Sorts the picture of the input video data to
Output to the scan conversion blocking unit 202.

The scan conversion block forming unit 202 performs field / frame conversion on the video data input from the picture rearranging unit 200 and further converts the video data into macroblocks to form the motion detecting unit 204 and the motion compensating unit 24 of the compression control unit 24.
Output for 0. The motion detection unit 204 converts the video data input from the scan conversion blocking unit 202 into a FIFO
206 and the subtraction circuit 242 of the compression control unit 24. The video data input from the scan conversion blocking unit 202 is processed in units of macroblocks to detect the motion, and a motion vector indicating the motion of the video. Is generated and output to the motion compensation unit 240 and the FIFO 220.

The FIFO 206 buffers the video data input from the motion detection unit 204, and stores only the time required for processing (preliminary coding processing) in the DCT unit 244, prediction unit 246 and back search unit 248 of the compression control unit 24. The output is provided to the subtraction circuit 207 with a delay.

The subtraction circuit 207 outputs, to the DCT unit 208, the video data of a picture that becomes an I-picture after compression encoding among the pictures included in the video data input from the FIFO 206. Further, the subtraction circuit 207 subtracts the output video data of the motion compensation unit 222 from the video data of a picture that becomes a P picture or a B picture after compression encoding among the pictures included in the video data input from the FIFO 206 and performs prediction. Error data is generated and output to DCT section 208.

The DCT section 208 performs DCT processing on the video data which is input from the subtraction circuit 207 to become an I picture after compression encoding and the prediction error data of the video data which becomes a P picture or a B picture after compression encoding. The DCT coefficient obtained as a result of the DCT processing is output to quantization section 210. The quantization unit 210 performs the quantization step indicated by the quantization index input from the back search unit 248 of the compression control unit 24,
8 is quantized and output to the variable length coding unit 212 and the inverse quantization unit 214 as quantized data.

The variable length coding section 212 includes a quantization section 210
Quantized data input from the VGA is variable-length encoded by, for example, a run-length encoding method, and output video data (strea
m out). The inverse quantization unit 214 performs a process reverse to that of the quantization unit 210 on the input quantized data to reproduce a DCT coefficient, and outputs the DCT coefficient to the inverse DCT unit 216.

The inverse DCT section 216 performs the inverse processing of the DCT section 208 on the input DCT coefficients to reproduce video data, and outputs the video data to the addition circuit 218. The addition circuit 218 adds up the video data input from the inverse DCT unit 216 and the video data input from the motion compensation unit 222 to reproduce the video data, and outputs the video data to the motion compensation unit 222.

The FIFO 220 buffers the motion vector input from the motion detection unit 204, and
206 to output to the motion compensator 222 with a delay by the time required for the processing of the adder circuit 218. Motion compensator 22
2 performs a motion compensation process on the video data input from the addition circuit 218 using the motion vector input from the FIFO 220, and outputs the result to the subtraction circuit 207 and the DCT unit 208.

The compression control unit 24 detects the quantization step and the GOP phase in the previous compression encoding of the input video data, and controls the compression encoding unit 20 so that they match. Further, the compression control unit 24 controls the complexity and the speed of movement of the pattern of the input video data [difficulty of the pattern (difficulty);
y], the total amount (data rate) of the output video data output from the variable length coding unit 212 is 1 to several GOs
Calculate a quantization step that is less than or equal to the permissible value (for example, the transmission capacity of the transmission path) of the output video data for each time (unit period) for generating the compressed video data for P; The value is set in the quantization unit 210 in the form of a quantization index.

In the compression controller 24, the motion compensator 24
0 is the same as the motion compensator 222 of the compression encoder 20.
The video data input from the scan conversion blocking unit 202 is subjected to a motion compensation process using the motion vector input from the motion detection unit 204, and is output to the subtraction circuit 242. The subtraction circuit 242 subtracts the motion-compensated video data input from the motion compensation unit 240 from the video data input from the motion detection unit 204, similarly to the subtraction circuit 207 of the compression encoding unit 20. , And the prediction error data of the P picture or the B picture are generated and output to the DCT section 244.

The DCT unit 244 outputs the D
Similarly to the CT unit 208, the I-picture video data and the P-picture or B-picture prediction error data input from the subtraction circuit 242 are DCT-
A DCT coefficient obtained as a result of the T processing is generated and output to the prediction unit 246.

The prediction section 246 outputs the DCT coefficient input from the DCT section 244 to the back search section 248. For example, the prediction unit 246 sets the time for generating compressed video data for one to several GOPs as a unit period, and sets a fixed value quantization step (fix-q) for each unit period.
The DCT coefficient input from the CT unit 244 is quantized to generate quantized data.

The prediction section 246 is set from the outside by a user of the video data processing system 1 (encoder 2), and sets a target data amount indicating the allowable value and a data amount of the generated quantized data. Based on the code amount, the difficulty (difficulty) of the pattern of the input video data is estimated for each unit period.

The predicting unit 246 allocates a large amount of data (data rate) to a portion where the pattern of the input video data is difficult according to the estimated difficulty of the pattern of the input video data. A small amount of data (data rate) is allocated to simple parts to keep the quality of the output video data high as a whole, and the quantization to be actually used so that the total amount of the output video data does not exceed the allowable value. A quantization index indicating a step is calculated for each unit period, and is output to the back search unit 248.

Note that, in addition to the method of predicting a quantization index by quantizing DCT coefficients in the above-described fixed-value quantization step, the prediction unit 246 includes, for example, DCT
The DCT coefficient input from the section 244 is quantized using some temporary quantized values to generate quantized data, and the data amount (generated code amount) of the generated quantized data and the target data amount (allowable value) ), It is also possible to adopt a method (binary search) of predicting an optimal quantization step by a binary tree search depending on which of these is larger.

Overview of Processing of Back Search Unit 248 The back search unit 248 performs the previous search based on the quantization index input from the prediction unit 246 and the DCT coefficient input from the DCT unit 244 via the prediction unit 246. , Detects an intra-coded picture in the compression coding of
Output to

The back search unit 248 determines whether or not the input video data has undergone one or more compression encodings by a back search, and determines the quantization step used in the previous compression encoding. An index is generated and set in the quantization unit 210 of the compression encoding unit 20.

That is, the back search section 248 divides the DCT coefficient input from the DCT section 244 via the prediction section 246 by the quantization step indicated by the quantization index predicted by the prediction section 246 and a value near the quantization step. If there is a quantization step in which the sum of the remainder of the division result indicates a significantly small value, the quantization step in which the significantly small value is determined is determined as the quantization step used in the previous compression encoding. A quantization index indicating a quantization step is output to quantization section 210.

The details of the processing of the back search section 248 Further, the contents of the processing of detecting the I picture of the input video data by the back search section 248 will be described in detail. The GOP phase detection process performed by the back search unit 248 is a process in which a picture decompressed and decoded from an I picture included in the input video data (also simply referred to as an “I picture of the input video data”) is again subjected to the I compression even in the next compression encoding. Only when the picture is compression-coded (intra-coded), attention is paid to the property that the residual sum of the DCT coefficients obtained by the back search shows a remarkable minimum.

The reason is that the P picture or the B picture is subjected to the DCT processing of the prediction error data calculated by the motion compensation processing and is compression-coded by quantization, so that the P picture or the B picture is inversely DC-coded.
Even if T processing, dequantization and motion compensation are performed for decompression decoding,
This is because the value of the video data obtained as a result of the decompression decoding cannot be an integral multiple of the quantization step. Therefore,
A picture obtained by decompressing and decoding a P picture or a B picture (hereinafter simply referred to as “P picture of input video data,
When the B-picture is also intra-coded and back-searched, there is no minimum point in the sum of the remainder of the DCT coefficients, and it cannot be found.

Similarly, when a B-picture (P-picture) of input video data is compression-encoded again into the same B-picture (P-picture) and a back search process is performed, the motion obtained by the previous compression-encoding is also obtained. The vector and the motion vector obtained from the video distorted by the next compression encoding are not the same, and because of the distortion of the video by the previous compression encoding and decompression decoding, in the next compression encoding, The obtained prediction error data is not the same as the prediction error data obtained in the previous compression encoding. Therefore, even if the B picture or P picture of the input video data is compressed and coded to the same picture type as the previous time, the probability of finding a minimum point at which the sum of the remainder of the DCT coefficient becomes a remarkably small value in the back search processing is small. Very low.

By utilizing the above-described nature of the back search processing for each picture type, it is determined whether or not the picture of the input video data has been previously intra-coded by the DCT coefficient in the back search processing. It can be automatically detected based on whether or not a local minimum occurs.

The back search unit 248 divides the DCT coefficient by a quantization step indicated by the quantization index input from the prediction unit 246 to determine whether or not there is a remarkable minimum value in the sum of the remainder of the DCT coefficient. The encoder 2 determines whether the ratio of the sum of the remainders of the DCT coefficients obtained by the back search processing in the back search unit 248 to the sum of the remainders in the case becomes equal to or smaller than a certain threshold. It can be determined whether or not the I picture of the input video data has been intra-coded.

As described above, the back search unit 248 detects the position of the I picture of the input video data, thereby determining the interval between the I pictures of the input video data (the number N of pictures included in the GOP). Yes, the determination result is notified to the picture type control unit 250.

The picture type control unit 250 determines the interval between the obtained I pictures and the configuration of the GOP used by the encoders 2a and 2b and the decoders 4a and 4b of the video data processing system 1 (what kind of picture type is included in the GOP). (GOP sequence), the interval (M) between P pictures in the GOP can be determined. Furthermore, each picture of the input video data Can be determined.

The picture type control section 250 controls each picture of the input video data based on the information indicating which picture type the input video data determined as described above was compression-coded in the previous compression coding. Is compressed and encoded to the same picture type as last time,
The picture rearranging unit 200 is controlled so that the GOP phase in the previous compression encoding and the next compression encoding is maintained, and the pictures are rearranged.

Operation of Encoder 2 The operation of the encoder 2 (FIGS. 1 and 2) shown in FIGS. 1 and 2 will be described with further reference to FIG. The motion compensation unit 240 of the compression control unit 24 includes the picture rearranging unit 200 and the scan conversion blocking unit 2 of the compression encoding unit 20.
02 performs motion compensation on the processed video data. Subtraction circuit 2
Reference numeral 42 generates video data of an I picture and prediction error data of a P picture or a B picture.
The DCT unit 244 performs DCT transform on the video data of the I picture and the prediction error data of the P picture or the B picture, and generates DCT coefficients.

The prediction unit 246 sets the time for generating compressed video data for one GOP as a unit period, and performs a DCT by a fixed quantization step (fix-q) for each unit period.
The DCT coefficient input from the unit 244 is quantized to generate quantized data, the difficulty of the input video data is estimated based on the generated code amount, and a quantization index is calculated for each unit period.

The back search section 248 detects a picture that has been intra-coded in the previous compression coding based on the quantization index generated by the prediction section 246 and the DCT coefficient generated by the DCT section 244. ,
A back search generates a quantization index indicating the quantization step used in the previous compression encoding. FIG. 3 shows the encoder 2 (2) shown in FIG. 1 and FIG.
FIGS. 7A and 7B are flowcharts illustrating the processing contents of the back search unit 248 and the picture type control unit 250 in a) and 2b), that is, how to specify a picture of input video data to be intra-coded.

As shown in FIG. 3, step 100 (S1
In (00), the back search unit 248 sets the variable j to the initial value 1. In step 102 (S102),
The back search unit 248 performs a back search on the DCT coefficient obtained as a result of the motion compensation unit 240, the subtraction circuit 242, and the DCT unit 244 of the compression encoding unit 20 compressing the j-th picture of the input video data. The processing is performed, and the sum R _{min, j} of the remainder of the DCT coefficients and the prediction unit 24
Quantization step 6 was obtained stores the sum R _j remainder of the DCT coefficients by (the previous quantization step).

In step 104 (S104), the back search unit 248 sets the variable j and the G
The number of pictures in the OP is compared with the number N of pictures. If the variable j is larger than the number N of pictures, the process proceeds to S108 and the variable j
Is less than or equal to the number N of pictures, the process proceeds to S106. In step 106 (S106), the back search unit 248 adds 1 to the variable j (increment)
I do.

In the loop processing composed of the processing of S100 to S106, the back search unit 248 sets N (1G
OP), the prediction unit 246 quantizes the DCT coefficient by a fixed-value quantization step, or the binary search using the sum R _{sub, j} of the remainder of the DCT coefficient obtained by the back search process. the resulting quantization step carries out a process of storing the remainder sum R _j when the quantized by.

In step 108 (S 108), the back search unit 248 calculates the residual sum R _j obtained by the quantization step obtained by the prediction unit 246 from S 100 to S 100.
N obtained by the loop processing constituted by the processing of S106
The minimum value [min ( _{Rmin, j} / _Rj )] of the ratio ( _{Rmin, j} / _Rj ) of the remainder sum _{Rsub, j} is compared with a predetermined threshold Th.

The back search unit 248 calculates the minimum value [min
(R _{min, j} / R _j )] is equal to or greater than the threshold Th,
When it is determined that the input video data is the original video data that has never undergone compression encoding, the process proceeds to S100, and when the minimum value [min (R _{min, j} / R _j )] is less than the threshold value Th In, it is determined that the input video data is video data that has already undergone compression encoding (after dubbing), and the process proceeds to S110.

In step 110 (S110), the back search unit 248 sets the j _min- th picture giving the minimum value [min (R _{min, j} / R _j )] to the I picture of the input video data in the processing of S108. And notifies the picture type control unit 250.
In step 112 (S112), the picture type control unit 250, the first j _min th picture, GO
Judge whether the picture is the first picture of P, and j
_{If the min-} th picture is the first picture of the GOP, the process proceeds to S100. If the j-th _min- th picture is not the first picture of the GOP, the process proceeds to S114.

In the processing of S112, the fact that the j _min- th picture is the first picture of the GOP means that the phase (configuration) of the GOP currently subjected to the back search processing is the previous one. Since this means that the phase (configuration) of the GOP targeted for the back search process is the same, the picture type control unit 250
There is no need to change the phase of the GOP.

On the other hand, in the processing of S112, the j-th
_{The fact that the min-} th picture is not the first picture of the GOP means that the phase (configuration) of the GOP currently being processed by the back search unit 248 is the same as that of the previous processing by the back search unit 248. GOP which had become
Therefore, the picture type control unit 250 needs to change the phase of the GOP.

[0082] At step 114 (S114), the picture type control unit 250, the N + j _min th input a picture of the video data so as to intra-picture reordering unit so as to change the order of the rearranged pictures 200 is controlled. Step 116
In (S116), the picture type control unit 250
Substitutes the numerical value j _min for the variable j, and proceeds to the process of S102.

The picture rearranging section 200 of the compression encoding section 20 includes the picture type control section 250 described above.
In the order of the picture of the input video data. The scan conversion blocking unit 202 performs field / frame conversion on the video data and further converts the video data into macroblocks.

The motion detecting section 204 processes the video data on a macroblock basis, detects the motion, and generates a motion vector indicating the motion of the video. The FIFO 206 buffers the video data and gives a predetermined time delay. The subtraction circuit 207 generates prediction error data of a picture that becomes a P picture or a B picture after compression encoding.

The DCT section 208 converts the prediction error data of the I-picture video data and the P-picture or B-picture video data after compression encoding into DC data.
T processing is performed to generate DCT coefficients. The quantization unit 210
In the quantization step indicated by the quantization index generated by the back search unit 248 of the compression control unit 24, the DCT unit 20
8 to quantize the DCT coefficient input thereto, and generate quantized data. The variable-length coding unit 212 includes a quantization unit 210
, Performs variable-length encoding on the quantized data input from, and outputs it as output video data (stream out).

The inverse quantization unit 214 performs an inverse quantization process on the quantized data to reproduce DCT coefficients. Inverse DCT section 216
Reproduces video data by performing inverse DCT processing on the reproduced DCT coefficients. The addition circuit 218 adds the video data input from the inverse DCT unit 216 and the video data input from the motion compensation unit 222.

The FIFO 220 buffers the motion vector input from the motion detector 204 and gives a predetermined time delay. The motion compensator 222 includes an adder 218
Is subjected to a motion compensation process using the motion vector input from the FIFO 220 to the video data input from the.

Modified Example Hereinafter, a modified example of the first embodiment will be described with reference to FIG. The encoder 2 (FIGS. 1 and 2) includes a compression encoding unit 2
When performing a back search process for obtaining a quantization index to be set in the quantization unit 210 of 0, the picture type in the previous compression encoding is detected, and the picture rearrangement process of the picture rearrangement unit 200 is controlled. With such a configuration, the amount of delay is large, and the order of rearranging pictures in the picture rearranging section 200 cannot be changed quickly.

However, by the back search process,
In order to find the minimum point of the residual sum of the DCT coefficients, it is not always necessary to calculate the residual sum of all the DCT coefficients of the picture; some macroblocks in the picture are extracted and the minimum point of the residual sum is calculated. It is sufficient to check for its presence.

A modified example (encoder 5) of the encoder 2 shown in the first embodiment described below focuses on such a point and is made to improve the processing speed of the encoder 2. , The function of generating the quantization index and the function of detecting the phase of the GOP
The input video data is compressed and coded after detecting the OP phase.

FIG. 4 is a diagram showing a configuration of an encoder 5 according to a modification of the first embodiment. In FIG. 4,
Among the components of the encoder 5, the same components as those of the encoder 2 are denoted by the same reference numerals as in FIG. As shown in FIG. 4, the encoder 5 includes a compression encoder 20 and a compression controller 2
6 is comprised. The compression control unit 26 includes the compression control unit 24
A configuration in which a GOP phase control unit 28 is added to FIG. 2 is adopted. The GOP phase control unit 28 includes a block extraction unit 280,
DCT section 282, prediction section 284, and back search section 2
86.

In the GOP phase control unit 28 of the compression control unit 26, the block extraction unit 280 extracts several blocks of video data from each picture of the input video data, and
Output to the CT unit 282. The DCT unit 282 performs DCT processing on video data of several blocks input from the block extraction unit 280, and obtains a DCT obtained by the DCT processing.
The T coefficient is output to the prediction unit 284.

The prediction unit 284, like the prediction unit 246 of the compression control unit 24, estimates the difficulty of the picture of the input video data by a fixed value quantization step (fix-q) or binary search for each unit period. A quantization index indicating a quantization step to be actually used is calculated for each unit period and output to the back search unit 286.

[0094] The back search section 286, like the back search section 248 of the compression control section 24 in the encoder 2, has the quantization index input from the prediction section 284,
Also, a back search process is performed using the DCT coefficient input from the DCT unit 282 via the prediction unit 284, and the GOP phase (I picture) of the input video data is detected.
The picture type control unit 250 is notified. The picture type control unit 250 controls the processing of the picture rearranging unit 200 of the compression encoding unit 20 according to the information indicating the position of the I picture input from the back search unit 286, as in the encoder 2.

Note that, unlike the encoder 2, the encoder 5 does not perform motion compensation prior to GOP phase detection (I-picture detection), so the back search unit 286 of the GOP phase control unit 28 The back search process is performed on the DCT coefficients obtained by intra-encoding all the pictures. Therefore, as a result, it is very easy for the GOP phase control unit 28 to determine whether or not the input video data has already been subjected to compression encoding.

Each component of the encoder 2 (2a, 2b) shown in FIG. 1 and FIG. 2 is configured by software or hardware as long as the same function and performance can be realized. It does not matter. Further, each component of the encoder 2 can be replaced with another device that can realize the same function and performance. The encoder 2 can be applied to compression encoding of other types of data having redundancy, such as audio data, in addition to video data, by appropriately modifying the encoder.

[0097] As the effect described above, according to the encoder 2 and 5, it is possible to perform the next compression encoding while maintaining the same GOP phase the previous compression-encoded, the video data processing system shown in FIG. 1 1 can be prevented from deteriorating the image quality.
In the encoders 2 and 5, when picture type information is lost, or when original video data that has never undergone compression encoding is input,
The compressed video data and the information indicating the picture type are multiplexed, and a malfunction such as a malfunction that occurs in the conventional method of performing the next compression encoding based on the picture type information does not occur.

In the encoders 2 and 5, for example, the compressed video data is returned to an analog video signal, and
Even when recording as digital compressed video data, compression encoding can be performed at the same GOP phase as the compressed video data before returning to analog video data, and deterioration of video quality can be prevented. Also, the encoder 5
Since the GOP phase is detected by using a part of macroblocks of the picture, the device scale does not greatly increase as compared with the encoder 2. Also, according to the encoder 5,
Since the processing time required for detecting the GOP phase is reduced, the delay time can be reduced as compared with the encoder 2.

Second Embodiment Hereinafter, a second embodiment of the present invention will be described.

Background of the Second Embodiment When the compression coding technique such as the MC-DCT method such as the MPEG2 method is used for broadcasting, communication or transmission of video data, the compression coding technique of the compressed video data obtained by the compression coding is used. In order to suppress the amount of data (the amount of generated code) to be equal to or less than the transmission capacity of the transmission path, the quantization step is usually controlled by feedback, as represented by TM5 of the MPEG system.

The control of the quantization step based on the feedback is performed by controlling the quantization step suitable for the buffer remaining amount at that time based on the relationship between the past quantization step (quantization index) and the generated code amount corresponding to the quantization step. This is performed by dynamically adjusting the quantization step (quantization index).

However, this feedback control method can suppress the amount of code generation to a desired data rate when viewed in a long time, but may increase the generated code amount when viewed instantaneously. Therefore, in consideration of application to a digital VTR device for a television broadcasting business or a recording / reproducing disk device, for example, concealment of insert editing, variable speed reproduction or recording error is considered.
al), the generated code amount is controlled in each unit period so that the generated code amount always reaches the target code amount, and is recorded in a specific location on a disk recording medium such as a VTR tape or MO disk. That control is further required.

For television broadcasting stations requiring high video quality, it is also strictly required to prevent deterioration of video quality when compression encoding and decompression decoding are repeated by tandem-connected encoders and decoders (codecs). Required.

Therefore, when the compression encoding and decompression decoding are repeated in the video data processing system 1 (FIG. 1) and the like, the encoders 2 and 5 are used to make the quantization steps in each compression encoding process the same. The back search processing by the back search unit 248 of FIGS. 1, 2, and 4 is used.

However, in the back search processing of the back search unit 248 of the encoders 2 and 5, the DCT
Since a method of detecting the quantization step used in the previous compression encoding based on whether or not there is a remarkable minimum point having a ratio exceeding a predetermined threshold value in the residual sum of the coefficients is adopted, If the threshold value is set too small, the minimum sum of the remainder of the DCT coefficients is reduced from some macroblocks of the picture of the original video data which have not undergone a single compression encoding, since there is no minimum point of the remainder sum of the DCT coefficients. Points may be detected incorrectly.

When the back search unit 248 erroneously detects the remainder sum of DCT coefficients and the quantization unit 210 of the compression encoding unit 20 performs quantization based on the erroneously generated quantization step (quantization index). However, there is a possibility that the DCT coefficient is quantized with a value larger than the optimal quantization step, and as a result, the quality of the image is largely deteriorated.

Conversely, if the threshold value used for detecting the DCT coefficient remainder sum is set too large, even if video data that has already undergone compression encoding is input, some macroblocks in the picture may have DCT coefficients. May not be able to detect the minimum point of the remainder sum of Originally, if the minimum point to be detected cannot be detected, a quantization process in the next compression encoding is performed in a quantization step different from that in the previous compression encoding, and the quality of the video is greatly deteriorated. From these viewpoints, it is necessary to appropriately select the threshold value used for detecting the residual sum of the DCT coefficients, but it is difficult to solve the above-mentioned problems, and the above problems are remarkable depending on the pattern of the input video data. May also appear.

Further, as an apparatus using the MPEG 4: 2: 2 Profile system, for example, there is a system in which a GOP has a two-frame structure of an I picture and a B picture, and an apparatus using the MP @ ML system. Is a GOP
Typically employs a 15-frame configuration.

On the other hand, in an MC-DCT encoder and decoder using a GOP having a plurality of frames, it is impossible to reproduce the same motion vector as the previous one in a picture type for which motion compensation processing is performed during compression encoding. Therefore, the reproducibility of the motion vector and the prediction error in the next compression encoding is low, and as a result, the residual sum of the DCT coefficients of the pictures compressed and encoded into the B picture and the P picture in the previous compression encoding is remarkable. There is no minimum point.

Therefore, even if the back search process is performed by the back search unit 248 using a threshold value for detecting a picture intra-coded in the previous compression encoding, the B picture and the P picture From a picture that has been compressed and encoded into
The minimum point of the CT coefficient cannot be detected. In the second embodiment,
To solve such a problem and realize a back search process that can prevent erroneous detection of picture type,
For example, an object of the present invention is to prevent deterioration of video quality when a plurality of digital VTR devices are connected to copy video data.

FIG. 5 is a diagram showing a configuration of the encoder 6 according to the present invention in the second embodiment. In FIG. 5, the same components as those of the encoder 2 shown in FIG. 2 and the encoder 5 shown in FIG. 4 among the components of the encoder 6 are denoted by the same reference numerals.

As shown in FIG. 5, the encoder 6 comprises a compression encoding section 20, a compression control section 30, and a video index (temporary name) detection section 32. The compression control section 30 includes encoders 2, 5 (Fig. 1, Fig. 2, Fig. 4)
And a switch (sw) circuit 300 is added to the compression control unit 24 of FIG. Note that the encoder 6 shown in the second embodiment is configured to perform motion compensation in compression encoding, but the motion compensation processing is not essential.

The encoder 6 is used, for example, in the video data processing system 1 (FIG. 1) in place of the encoders 2 and 5, and like the encoders 2 and 5, each picture contained in the input video data is Is detected, and the DCT coefficient is quantized in the same quantization step as in the previous time to perform compression encoding.

Back search processing in the encoder 6 In the back search processing in the encoder 6, it is better not to use the back search algorithm for the original video that has never undergone compression encoding, and to use an encoder that uses the same compression encoding method. Has been changed with a view to the property that it is preferable to apply only to video data that has already undergone compression encoding according to.

That is, in the encoders 2 and 5 shown in the first embodiment, the back search processing is performed on all the pictures of the input video data, and the minimum sum is added to the residual sum of the DCT coefficients in the original video data. Is not found, and it is expected that the minimum point will be found in the residual sum only in the video data that has already been subjected to the compression encoding, and that the back search processing will be effective. However, this may not be as expected.

Therefore, the encoder 6 shown in the second embodiment uses the video index to identify whether the picture of the input video data is video data that has already undergone compression encoding or original video data, and If the video data, and the previous time, if it is identified that the video data is compression-encoded by another method, the back search process is not performed, conversely, the previous time, compression encoding by the same method, In addition, the back search process is performed only when the previous GOP phase is the same as the GOP movement in the next compression encoding.

[0117] Detection method of the previous compression coding method Next, the previous compression coding method, a method of detecting whether the same or not the next compression coding method. Currently, SMP
In TE, it is being standardized that information indicating an encoding condition called a video index is multiplexed on video data on a decoder side. By detecting whether or not the correct video index is multiplexed in the video data, the input portion of the encoder 6 determines whether or not the input video data has been compression-coded in the same manner in the previous time, and It is possible to determine whether or not the phase (structure) of the GOP in the encoding and the phase of the GOP in the previous compression encoding match. Hereinafter, in the video data processing system 1, the decoders 4a and 4b
Multiplexes a video index with video data.

Overview of Operation of Video Index Detection Unit 32
In brief, an outline of the operation of the video index detection unit 32 will be described below. The video index detection unit 32 monitors whether or not the correct video index information is multiplexed on the input video data, and determines whether the correct video index information is multiplexed on the input video data, and Only when the phase (structure) of the GOP matches the phase of the GOP in the previous compression encoding, the phase (structure) of the GOP in the next compression encoding and the phase of the GOP in the previous compression encoding are changed. The operation of the picture rearranging unit 200 is controlled via the picture type control unit 250 so that they match with each other.
By controlling 0, the input terminal b is selected, and the quantization step (quantization index) determined by the processing of the back search unit 248 of the compression control unit 24 is output to the quantization unit 210.

The video index detecting section 32
Conversely, even if the video index is not multiplexed at all with the input video data, indicates a different compression coding method even if multiplexed, or indicates the same compression coding method, editing work etc. The GOP phase in the next compression encoding
If compression coding should not be performed in the OP phase, the switch circuit 300 is controlled to select the input terminal a, and the prediction unit 246 performs a fixed quantization step or a quantization step (index) obtained by a binary search. Is output to the quantization unit 210.

Details of operation of video index detecting section 32
Fine below with further reference to FIG. 7, a detailed operation of video index detection unit 32. FIG. 7 is a flowchart illustrating the operation of the video index detection unit 32 of the encoder 6 shown in FIG. 5. The video index detection unit 32 captures video data of one picture and then generates a quantization index. Are exemplified.

As shown in FIG. 7, step 200 (S2
In (00), the video index detection unit 32 of the encoder 6 captures the next input video data for one picture. In step 202 (S202), the video index detection unit 32 determines whether or not the next input video data has been captured. If the input video data has not been captured, the process ends. Step 204 (S20
In 4), the video index detection unit 32 separates and reads out all data (for example, 3 bytes) at a position where the video index of the input video data is to be multiplexed.

In step 206 (S206), the video index detection unit 32 performs a CRC check on the read 3-byte data. As a result of the CRC check, if the 3-byte data is correct, the process proceeds to S208; otherwise, the process proceeds to S216.

In step 208 (S208), the video index detecting unit 32 determines the GOP structure, the P picture interval (M) and the quantization method (Q_type) included in the video index based on the data having no change for each picture. Then, it is determined whether or not these conditions in the next (current) compression encoding match these conditions in the previous compression encoding. If they match, the process proceeds to S210; if they do not match, S2
Proceed to step 16.

In step 210 (S210), the video index detecting section 32 expects the picture type (picture type) and the number of frames (N; frame number) included in the index in the next (current) compression encoding. The picture type and the number of frames (N) are compared to determine whether or not they match. If they match, the process proceeds to S212, and if they do not match, the process proceeds to S216.

That is, the video index detecting section 32
Checks whether or not the data indicating the picture type and the data indicating the frame number included in the video index match the expected pattern. For example,
When the GOP has a two-frame configuration of one B picture and one I picture, GO
The first P is expected to be a B picture and the second P is expected to be an I picture.

Therefore, the video index detecting section 32
Is when the data (Frame No.) indicating the frame number included in the video index data is 1 and the data indicating the picture type included in the video index data is a B picture, and Data indicating the included frame number (F
It is determined that the video index is correct only in the two cases where the “rame No.” is 2 and the data indicating the picture type included in the video index data is an I picture.

In step 212 (S212), the video index detecting section 32 determines whether or not the information indicating the GOP phase included in the index matches the GOP phase in the next (current) compression encoding. .
If they match, the process proceeds to S214. If they do not match, the process proceeds to S216.

That is, the video index detecting section 32
Is the GOP phase in the next (current) compression encoding,
Only when the data indicating the GOP phase included in the video index matches, the quantization index generated by the back search unit 248 is validated.

In step 214 (S 214), the video index detection unit 32 controls the switch circuit 300 to select the input terminal b side, and the back search unit 248 generates the quantization unit 210 of the compression encoding unit 20. Output the quantization index.

That is, the video index detecting section 32
Means that the back search unit 248 is used only when all the conditions shown in the processing of S202 and S206 to S212 are satisfied.
Is enabled, the switch circuit 300 is controlled in the process of S214, and the quantization index generated by the back search unit 248 is quantized by the quantization unit 210.
Output.

In step 216 (S 216), the video index detecting section 32 causes the switch circuit 300 to select the input terminal “a”, and the quantizing section 210 makes a decision instead of the quantization index generated by the back search section 248.
Control is performed so that quantization is performed using the quantization index generated by the prediction unit 246 (free run processing). That is, the video index detection unit 32 determines that S2
02, if any of the conditions shown in the processing of S206 to S212 is not satisfied, the back search processing in the back search unit 248 is invalidated, the switch circuit 300 is controlled in S216, and the quantization generated by the prediction unit 246 is performed. The index is output to the quantization unit 210.

Decoder 4 (4a, 4b) FIG. 6 is a diagram showing the structure of the decoder 4 (4a, 4b) shown in FIG. When the encoder 6 is used instead of the encoders 2 and 5 in the video data processing system 1 (FIG. 1), the decoders 4a and 4b
A video index multiplexing unit 414 is added to the decompression decoding unit 40.
Is adopted.

The decompression / decoding section 40 has a buffer memory (buffer) 40 in the same manner as a general video data decoder.
0, variable length decoding unit (VLD) 402, inverse quantization unit 40
4, an inverse DCT unit 406, a motion compensation unit 408, a switch circuit 410, a picture rearrangement unit 412, and a picture type control unit 416.

The decompression decoding unit 40 decompresses and decodes the compressed video data input from the encoder 2, generates full-bit video data, and outputs it to the video index multiplexing unit 414. Video index multiplexing section 414
Generates a video index indicating the previous compression encoding method and the GOP configuration based on the GOP sequence of the compressed video data detected by the picture type control unit 416, and outputs the video index input from the picture rearranging unit 412. , And outputs the result to the encoder 2.

The video data processing system according to the second embodiment
Following the operation of the stem 1 , the encoder 6 (FIG. 5) and the decoder 4 (FIG. 6)
The operation of the video data processing system 1 (FIG. 1) using the above will be described. The encoder 6 (6a) compresses and encodes the input video data and transmits it to the decoder 4 (4a) via the recording / transmission device 3a. The decoder 4a expands and decodes the compressed video data input from the encoder 6a to generate full-bit video data, multiplexes the video index data, and transmits the multiplexed video index data to the encoder 6 (6b) via the recording / transmission device 3b. I do.

In the encoder 6b, the video index detecting section 32 (FIG. 5) performs the processing shown in FIG. 7 every time video data of one picture is input, and multiplexes the video data with the input video data. It is determined whether or not the video index is correct, and the switch circuit 300 is controlled so that the quantization step (quantization index) generated by the back search section 248 and the prediction section 246 sets any of the quantization steps (quantization indexes) generated, controls the picture rearranging unit 200 of the compression encoding unit 20 via the picture type control unit 250, and performs compression encoding in the encoder 6b. The pictures of the input video data are rearranged in a suitable order.

The components of the compression encoding unit 20 after the picture rearranging unit 200 and the compression control unit 30
Like the encoders 2 and 5, the input video data is compression-encoded and transmitted to the decoder 4b via the recording / transmission device 3c.

When the picture rearranging section 200 of the compression encoding section 20 is controlled based on the video index as in the encoder 6, the picture type at the time of the previous compression encoding is also used in the next compression encoding. Based on the picture type at the previous compression encoding, the threshold used in the back search processing of the back search unit 248 of the compression control unit 30 in the next and subsequent compression encodings is optimized according to the picture type. It is possible to

That is, when a picture that has been intra-encoded during the previous compression encoding is also intra-encoded next time, a remarkable minimum point of the remainder sum of the DCT coefficients tends to appear in the back search processing (FIG. 3). . Therefore, erroneous detection of the minimum point of the residual sum of the DCT coefficients can be prevented by setting the threshold value used for the back search process to be relatively large.

Conversely, as described above, when the pictures that have been compressed and coded into the B picture and the P picture in the previous compression coding are to be compression coded to the same picture type in the next compression coding, Because of the compression coding using motion prediction, the motion vector obtained in the previous compression coding does not match the motion vector obtained from the distorted video data from the next time, and the P picture and the B picture Even if the residual sum of the DCT coefficients obtained by performing the DCT processing on the prediction error is obtained, the minimum point does not appear so remarkably.

For this reason, in the case where a picture that has been compression-coded into a B picture and a P picture in the previous compression coding is to be compression-coded to the same picture type in the next compression coding, the back search unit 248 is used.
If the threshold value used in the backser process is set higher, there is a possibility that the back search unit 248 cannot find a correct quantization step (quantization index).

Therefore, when the pictures which have been compressed and coded into the B picture and the P picture in the previous compression coding are to be compressed and coded to the same picture type in the next compression coding, the back search unit 248 By setting the threshold used in the processing to a value smaller than that in the case of performing the above-described intra coding, the back search unit 248 can obtain a correct quantization step (quantization index).

Modification In the encoder 6 shown in FIG. 5, the video index detection unit 32 controls the switch circuit 300 to change the quantization step (quantization index). The operation is changed so that the operation of the back search unit 248 is directly turned on / off. When the operation of the back search unit 248 is turned off by the video index detection unit 32, the operation is input from the prediction unit 246. The quantization step (quantization index) is output to the quantization unit 210, and the operation is
When N is set to N, the quantization step (quantization index) generated by the back search unit 248 itself is
The output may be changed so as to be output to 10. Also, the encoder 6 can be changed in the same manner as the encoders 2 and 5 (FIGS. 1, 2 and 4).

[0144] As the effect described above, according to the encoder 6 shown in the second embodiment, when the video index detection unit 32 of the compression encoding section 30, the original image data is input, the back Since the back search processing of the search unit 248 is invalidated, it is possible to prevent a malfunction such as compression encoding of the original video data using a large quantization step, and to obtain a compression obtained by compression encoding the original video data. The image quality of the image data is improved.

In the encoder 6, the video index detection unit 32 of the compression encoding unit 30 invalidates the back search processing of the back search unit 248 when the original video data is input. When video data that has already been compression-encoded is input, the value of the threshold used by the back search unit 248 can be optimized to be small, and detection of a minimum point of the residual sum of DCT coefficients can be prevented from being missed. The system of back search processing is improved. Therefore, as a result, according to the encoder 6, the quality of the compressed video data obtained by compressing and encoding the video data that has already been subjected to the compression encoding is also improved.

Third Embodiment Hereinafter, a third embodiment of the present invention will be described.

Background of Third Embodiment FIG. 8 is a diagram showing a configuration example of the video data compression / multiplexing device 7. For example, a plurality of compressed video data may be multiplexed between television broadcast stations and transmitted via a communication line. In such a case, for example, the video data compression / multiplexing device 7 illustrated in FIG. 8 is used.

As shown in FIG. 8, the video data compression / multiplexing device 7 is composed of three encoders 70a to 70c and a multiplexing device 72. In the video data compression / multiplexing device 7, the encoders 70a to 70c respectively compress and encode input video data CH1 to CH3 input from, for example, a digital VTR device, and compress video data CH1 'at fixed data rates FR1 to FR3, respectively.
ＣＨCH3 'to the multiplexer 72. The multiplexing device 72 multiplexes the compressed video data CH1 ′ to CH3 ′ input from the encoders 70a to 70c, and outputs the multiplexed video data to a communication line (not shown) as output video data at an output data rate T.

Since the transmission capacity of the communication line connected to the multiplexing device 72 is predetermined, the output data rate T of the multiplexing device 72 is determined by the transmission capacity (allowable value) of this communication line.
Limited to: Therefore, the user of the video data compression / multiplexing device 7 distributes the output data rate of the multiplexing device 72 to each of the encoders 70a to 70c, and sets the data rates FR1 to FR3 (F
R1 + FR2 + FR3 ≦ T) must be set.

In such a case, simply the encoder 7
In addition to the method of setting a data rate of 1/3 of the output data rate T to each of the input video data C
A method of allocating the output data rate T according to the difficulty of H1 to CH3 may be adopted, and the latter allocation method is called statistical multiplexing.

For example, if the output data rate T of the multiplexer 72 is 10 Mbps, and the input video data CH1 is
If the input video data CH2 is video data of a news video having a relatively difficult pattern, and the input video data CH3 is video data of a movie having a relatively easy pattern, The user allocates 5 Mbps as the data rate FR1 to the encoder 70a, allocates 3 Mbps as the data rate FR2 to the encoder 70b, and allocates 2 Mbps as the data rate FR3 to the encoder 70c according to the statistical multiplexing method.
bps is allocated and set for each.

However, the difficulty of the picture of the video data changes with time, and the picture of the input video data CH3 may be much more difficult than the picture of the input video data CH1. In such a case, according to the video data compression / multiplexing device 7, since the data rates FR1 to FR3 are fixedly set for the encoders 70a to 70c,
The quality of the video of the compressed video data CH3 'is significantly deteriorated. The third embodiment has been made to solve such a problem.

Configuration of Video Data Compression / Multiplexing Apparatus 8 FIG. 9 is a diagram showing the configuration of the video data compression / multiplexing apparatus 8 according to the third embodiment of the present invention. An example in which the device 8 compression-encodes and multiplexes three pieces of input video data will be described. It should be noted that, among the components of the video data compression / multiplexing device 8, the encoders 2,
5 (FIGS. 1, 2 and 4) and the video data compression / multiplexing device 7 (FIG. 7) are denoted by the same reference numerals.

As shown in FIG. 9, the video data compression / multiplexing device 8 includes encoders 80a to 80c having the same configuration as any of the encoders 2, 5, and 6, a multiplexing device 72, and a control unit (CPU) 82. Is done. In the video data compression / multiplexing device 8, the compression control units 24 (26, 30, hereinafter simply referred to as the compression control units 2) of the encoders 80a to 80c.
The prediction unit 246 outputs the generated code amount (difficulty) of the input video data CH1 to CH3 obtained for each unit period by the fixed-value quantization step to the control unit 82. The operation has been changed so that the quantization step for the quantization unit 210 is calculated based on the target data amount input for each unit period from, and the operation is disabled even if the back search unit 248 operates. May be.

Overview of Operation of Control Unit 82 The control unit 82 determines the degree of difficulty of input video data input from the encoders 80a to 80c for each unit period (for example, the time for generating compressed video data for one GOP). The value of the data rate to be distributed to each of the encoders 80a to 80c is calculated and set in the prediction unit 246 of the compression encoding unit 24 of the encoders 80a to 80c, and the compressed video data C output by each of the encoders 80a to 80c is calculated.
The data rates E1 to E3 of H1 ′ to CH3 ′ are dynamically adjusted for each unit period.

Operation of Video Data Compression / Multiplexing Apparatus 8 The operation of the video data compression / multiplexing apparatus 8 will be further described below. In each of the encoders 80a to 80c,
The prediction unit 246 of the compression control unit 24 determines the unit period (1 GO
The difficulty levels D1 to D3 of the input video data CH1 to CH3 for P) are output to the control unit 82.

The control unit 82 calculates the following equation 1-1 or equation 1-
2, the total allowable generated code amount P for each unit period is calculated.

[0158]

## EQU1 ## P = [T (bps) × N (sheet)] / 30 (frame) = TN / 30 (bit) (in the case of NTSC; 1-1) P = [T (bps) × N (sheet) )] / 25 (frame) = TN / 25 (bits) (in the case of PAL; 1-2) where N is the number of pictures included in a unit period (1 GOP), and T is video data compression / multiplexing. The data rate allowed by the communication line or the like connected to the device 8.

Further, control section 82 determines the difficulty levels D1 to D1 of input video data CH1 to CH3 input from prediction section 246.
Using D3, as shown in the following equations 2-1 to 2-3, equations 3-1 to 3-3, the total allowable generated code amount P is, for example, proportionally distributed, and the target code amounts E1 to E3 are calculated. It is calculated and set in the prediction unit 246 of the compression control unit 24.

[0160]

E1 = (TN / 30) × [D1 / (D1 + D2 + D3)] (bit) (2-1) E2 = (TN / 30) × [D2 / (D1 + D2 + D3)] (bit) (2) 2) E3 = (TN / 30) × [D3 / (D1 + D2 + D3)] (bits) (2-3) where the input video data is NTS
It is suitable for the case of the C method, and the decimal places of E1 to E3 are truncated.

[0161]

E1 = (TN / 25) × [D1 / (D1 + D2 + D3)] (bit) (3-1) E2 = (TN / 25) × [D2 / (D1 + D2 + D3)] (bit) (3- 2) E3 = (TN / 25) × [D3 / (D1 + D2 + D3)] (bits) (3-3) However, in Equations 3-1 to 3-3, the input video data is PAL.
Compatible with the method.

It should be noted that the video data compression / multiplexing device 8
In general, the case of processing input video data of
The target data amount Ei allocated to each of the encoders 80a to 80n is expressed by the following equations 4-1 and 4-2.

[0163]

Ei = (TN / 30) × [Di / (D1 +... + Dn)] (bit) (in the case of NTSC; 4-1) Ei = (TN / 25) × [D2 / (D1 + D2 + D3)] ( (Bit)... (In the case of PAL; 4-2) where i = 1 to n, and the fractional part of Ei is rounded down.

Specific examples will be described. For example, the input video data is in the NTSC format, N = 30, T = 10 (Mbp)
s), D1 = 10 (Mbit), D2 = 20 (Mbi)
t), when D3 = 30 (Mbit), the control unit 82 determines the difficulty levels D1 to D3 of the input video data CH1 to CH3.
3, the target data amounts E1 to E3 are calculated by the following equation 5-1.
To 5-3.

[0165]

E1 = [(10 × 30) / 30] × [10 / (10 + 20 + 30)] = about 1.6 Mbps (5-1) E2 = [(10 × 30) / 30] × [20 / ( 10 + 20 + 30)] = about 3.3 Mbps (5-2) E3 = [(10 × 30) / 30] × [30 / (10 + 20 + 30)] = about 5.0 Mbps (5-3)

The prediction unit 246 of each of the encoders 80a to 80c calculates a quantization step (quantization index) based on the target data amount set by the back search unit 248, and calculates the quantization step (quantization index) of the quantization unit 210 of the compression encoding unit 20. Set to. The compression encoding unit 20 of each of the encoders 80a to 80c performs the compression encoding using the quantization step (quantization index) set by the back search unit 248, and the target data amount E1 to E3 or less per unit period. Thus, the compressed video data CH1 ′ to CH3 ′ having a data amount substantially close to the target data amounts E1 to E3 is generated and output to the multiplexer 72. The multiplexer 72 multiplexes the compressed video data CH1 ′ to CH3 ′ input from the encoders 80a to 80c to generate output video data, and outputs the output video data to a communication line (not shown).

Modification 1 Hereinafter, a first modification of the operation of the video data compression / multiplexing device 8 described in the third embodiment will be described. For example, if one of the video data input to the video data compression / multiplexing device 8 is video data such as a sports program in which a pattern is difficult, the video data is transmitted through the entire video data.
There are cases where it is desired to perform compression encoding at a high data rate.

In such a case, the encoders 80a to 80a
80c, the difficulty level D1 input from each prediction unit 246.
ＤD3 (generalized as Dk (k = 1 to n)) is multiplied by a weighting coefficient Ak to give a priority (weight)
And the operation of the control unit 82 may be modified so as to calculate the target data amount Ek as shown in Equations 6-1 and 6-2 below.

[0169]

Ek = (TN / 30) × [Ak × Dk / (A1 × D1 +... + An × Dn)] (in the case of NTSC; 6-1) Ek = (TN / 25) × [Ak × Dk / (A1 × D1 +... + An × Dn)] (for PAL; 6-2)

Specific examples will be described. Input video data is NTS
C method, N = 30 sheets, T = 10 (Mbps), D
1 = 10 (Mbit), D2 = 20 (Mbit), D3
= 30 (Mbit) and the input video data CH1 to C
The weighting coefficients A1 to A3 for H3 are 1,
In the case of 3, 2, the target data amounts E1 to E3 can be obtained as shown in the following equations 7-1 to 7-3.

[0171]

E1 = (10 × 30/30) × [(1 × 10) / (1 × 10 + 3 × 20 + 2 × 30)] = about 0.7 Mbps (7-1) E2 = (10 × 30/30) ) × [(3 × 20) / (1 × 10 + 3 × 20 + 2 × 30)] = about 4.6 Mbps (7-2) E3 = (10 × 30/30) × [(2 × 30) / (1 × 10 + 3 × 20 + 2 × 30)] = about 4.6 Mbps (7-3)

Modification 2 Hereinafter, a second modification of the video data compression / multiplexing device 8 shown in the third embodiment will be described with reference to FIG. FIG. 10 is a flowchart showing the operation of the second modification of the video data compression / multiplexing device 8 shown in FIG.

According to the capacity of the FIFO 206 of the compression encoder 20 of each of the encoders 80a to 80c, the GOP
Instead of the unit, it is possible to flexibly set the unit period to about several seconds from the time of generating the compressed video data for one frame.

As shown in FIG. 10, step 300 (S
300), the control unit 82 controls the encoders 80a to 80a.
80c calculated input video data CH1 to CH
3, the difficulty levels d1 to d3 for each frame are captured for each frame. In step 302 (S302), the control unit 82 cumulatively adds the difficulty levels d1 to d3 of the input video data (D1 = d1 ++, D2 = d2 ++, D3 = d3).
++).

In step 304 (S304), the control unit 82 determines whether the counted number of frames is N or not.
That is, in the processing of S302, the difficulty levels D1 to D1 for one unit period (time to generate compressed video data for N frames)
It is determined whether or not the accumulation of D3 has been completed. If the processing has been completed, the process proceeds to S306, and if not, the process returns to S302.

In step 306 (S 306), the control section 82 sets the formulas 2-1 to 2-3 (when the input video data is N
The target data amounts E1 to E3 are calculated according to the TSC method or the formulas 3-1 to 3-3 (when the input video data is the PAL method).

In step 308 (S308), the control unit 82 sends the calculated target data amount E to the prediction unit 246 of the compression control unit 24 of each of the encoders 80a to 80c.
1 to E3, and the data rate E1 of the compressed video data
Adjust ~ E3. Further, encoders 80a to 80c
Each of the compression encoding units 20 compression-encodes the input video data CH1 to CH3 using the set target data amounts E1 to E3, and the generated code amount in the unit period is equal to or less than the target data amounts E1 to E3. , Target data amount E1
It generates and outputs compressed video data CH1 'to CH3' substantially the same as E3.

Note that the number of encoders 80a to 80c of the video data compression / multiplexing device 8 is an example, and the number of input terminals of the multiplexing device 72 and the processing content of the control unit 82 can be changed as appropriate. Number.
Also, the same modification can be applied to the video data compression / multiplexing device 8 as to the encoders 2, 5, and 6.
Further, the video data compression / multiplexing devices 7 and 8 can be used in the video data processing system 1 instead of the encoders 2, 5 and 6.

Effects As described above in the third embodiment, according to the video data compression / multiplexing apparatus 8 according to the present invention, based on the degree of difficulty of the input video data obtained by the prediction section 246 of the compression control section 24. , The value of the output data rate allocated to each of the plurality of input video data can be dynamically adjusted, the output data rate of the multiplexer 72 can be used effectively, and the compressed video data CH1 ′ to CH1
The image quality of 3 'is improved as a whole.

According to the first modification of the video data compression / multiplexing device 8, the target data amount is calculated by weighting the degree of difficulty of the input video data, so that more effective compressed video data can be obtained. Multiplexing becomes possible. Further, according to the second modification of the video data compression / multiplexing device 8, the unit period for adjusting the target data amount is not set in GOP units,
It can be changed flexibly in units of one frame.

[0181]

As described above, according to the data compression apparatus and method according to the present invention, the picture data at the time of the previous compression encoding can be obtained without specially multiplexing the picture type information with the effective pixel data. The type can be automatically detected on the encoder side, and the GOP phase can be matched and compression-encoded.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a video data processing system in which an encoder according to the present invention is used.

FIG. 2 is a diagram showing a configuration of an encoder according to the present invention in the first embodiment shown in FIG.

FIG. 3 is a flowchart illustrating processing contents of a back search unit and a picture type control unit of the encoder shown in FIGS. 1 and 2;

FIG. 4 is a diagram illustrating a configuration of an encoder according to a modified example of the first embodiment.

FIG. 5 is a diagram illustrating a configuration of an encoder according to the present invention in a second embodiment.

FIG. 6 is a diagram illustrating a configuration of a decoder illustrated in FIG. 1;

FIG. 7 is a flowchart illustrating an operation of a video index detection unit of the encoder shown in FIG. 5;

FIG. 8 is a diagram illustrating a configuration example of a video data compression / multiplexing device.

FIG. 9 is a diagram showing a configuration of a video data compression / multiplexing apparatus according to the present invention in a third embodiment.

FIG. 10 is a flowchart showing an operation of a second modification of the video data compression / multiplexing apparatus shown in FIG. 9;

[Explanation of symbols]

1. Video data processing system, 2 (2a, 2b), 5,
6 encoder, 20 compression encoding unit, 200 picture rearranging unit, 202 scan conversion blocking unit, 20
4: Motion detector, 206: FIFO, 208: DCT
Unit, 210: quantization unit, 212: variable length encoding unit, 21
4 Inverse quantization unit, 216 Inverse DCT unit, 218 Addition circuit, 220 FIFO, 222 Motion compensation unit, 24, 2
6, 30 ... compression control unit, 240 ... motion compensation unit, 242 ...
Subtraction circuit, 244... DCT section, 246... Prediction section, 248
... back search section, 250 ... picture type control section,
28: GOP phase control unit, 280: block extraction unit, 2
82 DCT unit, 284 prediction unit, 286 back search unit, 300 switch circuit, 32 video index detection unit, 3 (3a to 3c) recording / transmission device, 4 (4
a, 4b): decoder, 40: decompression decoder, 400: buffer memory, 402: variable length decoder, 404: inverse quantization unit, 406: inverse DCT unit, 408: motion compensation unit, 41
0: switch circuit, 412: picture rearranging section, 4
14 video index multiplexing section, 416 picture type control section, 7, 8 video data compression / multiplexing apparatus, 70a-70c, 80a-80c encoder, 7
2. Multiplexer, 82 Control unit.

Claims (10)

[Claims] 1. An I-picture for detecting a picture which has been compressed and coded into an I-picture in said compression processing from pictures of video data obtained by expanding compressed video data which has been compressed into a combination of an I picture, a P picture and a B picture. Each of the pictures of the video data is subjected to I-picture processing in the compression processing based on picture detection means and an interval between the detected I-pictures.
Picture type determining means for determining which picture type among pictures, P pictures and B pictures has been compressed; and video data compressing means for compressing each picture of the video data into the same picture type as in the compression processing. A video data compression device having: 2. An I-picture detecting means, comprising: a orthogonal transformation means for orthogonally transforming each picture of the video data for each macroblock to generate orthogonal conversion data; and for each macroblock of each picture of the generated video data. A dividing means for dividing the orthogonal transform data by each of a plurality of quantization steps; anda sum of remainders of respective division results of the orthogonal transform data by the plurality of quantization steps for each macroblock of each picture of the video data. A remainder sum calculating means for calculating; and a detecting means for detecting whether or not the picture of the video data is compression-encoded into an I-picture in the compression processing based on whether or not there is a minimum value of the sum of the remainders The video data compression device according to claim 1, comprising: 3. An I-picture detecting means, comprising: an orthogonal conversion means for orthogonally converting a part of a macroblock of each picture of the video data to generate orthogonal conversion data; and a macro for each picture of the generated video data picture. Division means for dividing the orthogonal transform data of a part of a block by each of a plurality of quantization steps; A residue sum calculating means for calculating the sum of the respective remainders, and whether or not the picture of the video data is compression-encoded into an I-picture in the compression process based on whether or not there is a minimum value of the sum of the remainders 2. The video data compression apparatus according to claim 1, further comprising: a detection unit configured to detect whether or not the video data is compressed. 4. The I-picture detecting means includes a quantization step selecting means for selecting a quantization step that minimizes a minimum value of the sum of the remainders, and the video data compressing means includes: 3. The video data compression apparatus according to claim 2, further comprising a quantization unit that quantizes the video data using a step. 5. The video picture according to claim 2, wherein said I picture detection means detects a picture which has been compression-encoded into an I picture in said compression processing before said video data compression means compresses said video data. Data compression device. 6. A picture which has been compressed and coded into an I picture in said compression processing among pictures of video data obtained by expanding compressed video data which has been compressed to a combination of an I picture, a P picture and a B picture, and Based on the detected I-picture intervals, each of the pictures of the video data is
A video data compression method for determining which picture type among a picture, a P picture and a B picture has been compressed, and compressing each picture of the video data into the same picture type as in the compression process. 7. A method for orthogonally transforming each picture of said video data for each macroblock to generate orthogonally-converted data, wherein said orthogonally-transformed data for each macroblock of each picture of said generated video data is subjected to a plurality of quantization steps. Calculating the sum of the respective remainders of the orthogonal transformation data for each macroblock of each picture of the video data by the plurality of quantization steps, and determining whether there is a minimum value of the sum of the remainders And detecting whether or not the picture of the video data has been compression-encoded into an I-picture in the compression processing.
2. The video data compression method according to 1. 8. A method for orthogonally transforming a part of a macroblock of each picture of said video data to generate orthogonally transformed data, wherein a plurality of said orthogonally transformed data of a part of the generated macroblock of each picture of said video data are provided. Calculating the sum of the remainders of the respective division results of the orthogonal transformation data of the part of the macroblocks of each picture macroblock of the picture data of the video data, and calculating the sum of the sums of the remainders 7. A detection as to whether a picture of the video data has been compression-encoded into an I-picture in the compression processing based on whether or not there is a minimum value.
2. The video data compression method according to 1. 9. The video data compression method according to claim 7, wherein a quantization step that minimizes the minimum value of the sum of the remainders is selected, and the video data is quantized using the selected quantization step. 10. The video data compression method according to claim 7, wherein a picture compressed and encoded into an I picture in said compression processing is detected before said video data is compressed.