JPH08336138A - Moving image compression device - Google Patents

Abstract

PURPOSE: To fit the minute and second of a time code to the boundary of an image aggregate. CONSTITUTION: A moving image signal is inputted from an image signal input terminal 1. While, when a drop frame is included in the time code inputted from a time code input terminal 101, it is detected by a drop frame predicting/ detecting means 103. A GOP constitution decision means 104 decides the length of a GOP and each frame setting which picture out of I, P and B pictures as an image compression is performed based on a detected result, and the compression is performed with decided GOP length and picture type by an encoder means 2, and the signal is outputted from a bit stream output terminal 3.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture compression apparatus for compressing a digital moving picture.

[0002]

2. Description of the Related Art MPEG (Moving Picture Experts Gro)
up) 1, In moving image compression by MPEG2, an image is divided into small blocks (macroblock = rectangle of 16 pixels × 16 pixels), and the temporally preceding and following images are similar to the macroblock to be compressed. A region (reference image region) is extracted, spatial distance and orientation (motion vector) with respect to the reference image region, and difference information between the reference image region and the region to be compressed are calculated, and these information are calculated by the DCT, It is compressed using variable length coding. Thus, compressing only the motion vector and the difference information is much more efficient than compressing the original image itself.

The picture compressed by the difference information cannot be restored unless there is a picture to refer to, so it is necessary to periodically provide pictures that do not refer to other pictures. This image is called an I (Intra) picture. I
The next image is compressed using the picture as the reference image, and the next image is further compressed using the already compressed image as the reference image. Among pictures to be compressed using reference images, a P (Predictive) picture in which only a temporally preceding picture is a reference image and a B (Bidirectionally-Predicti) in which temporally preceding and subsequent pictures are reference images are included.
ve) There is a picture. Like the I picture, the P picture can be a reference image of another image. The I picture requires a larger amount of information than the P and B pictures. An aggregate of one or more I pictures and P and B pictures compressed using the I pictures as a reference image is called a GOP (Group of Picture).

FIG. 7A shows an example of GOP. In this figure, an input image is compressed as an I picture at a cycle of 10 frames, 9 frames in the meantime are compressed as P pictures, and a B picture does not exist.
In the figure, N is the length of GOP, and M is the period of I or P picture.

In this example, frame 0 is first compressed as an I picture without referring to other images, and then,
Frame 1 is compressed as a P picture with reference to the previous frame 0. Subsequent P pictures are sequentially compressed using the previous frame as a reference image. In this configuration, one GOP is composed of 10 frames.

FIG. 7 (b) shows another example of the GOP. In this figure, the input image is compressed as I pictures at a cycle of 10 frames, compressed every other frame as P pictures, and the remaining 5 frames are compressed as B pictures.

In this example, first, frame 1 is compressed as an I picture without referring to other images, and then,
Frame 0 is compressed as a B picture with reference to the subsequent frame 1. Next, the frame 3 is compressed as a P picture by referring to the front frame 1, and the frame 2 is compressed as a B picture by using the preceding and following frames 1 and 3 as reference images.

FIG. 7 (c) shows another example of the GOP. In this figure, the input image is compressed as an I picture at a cycle of 6 frames, compressed every other frame as a P picture, and the remaining 4 frames as a B picture.

In this example, first, frame 2 is compressed as an I picture without referring to other images, and then,
Frames 0 and 1 are compressed as B pictures with reference to the subsequent frame 2. Next, the frame 5 is compressed as a P picture with reference to the front frame 2, and the frames 3 and 4 are compressed as B pictures and the preceding and following frames 2 and 5 are compressed as reference images.

The length of GOP can be arbitrarily determined. The shorter the GOP length, the greater the proportion of I pictures and the lower the compression efficiency. However, the more I-pictures there are, the more the signal obtained as a result of compression (bitstream)
The point that a reproduced image can be obtained increases from the middle of.
Therefore, the GOP length is generally determined by the usage of the bitstream. For example, when used for a video conference, it is better that the bit stream obtained as a result of compression is as small as possible, and even if the bit stream is interrupted, it is possible to issue a resend request from the receiving side to the transmitting side. The I-picture cycle may be long. On the other hand, when performing a television broadcast using a bitstream, it is necessary for the viewer to output a reproduced image in a relatively short time even if the viewer starts receiving from the middle of the bitstream as a result of switching channels. The shorter the picture interval, the better.

As described above, since the length of the GOP is set according to the use, the GOP is set in the middle of the bit stream.
There was no need to change the length of the. Therefore, in the conventional MPEG encoder, the GOP length remains the value set at the start of compression and is not changed until the end of the bitstream.

[0012]

At the present time, MPEG is used.
Encoders that perform image compression according to the two standards are expensive,
It is also inconvenient to carry. Therefore, when the MPEG2 standard is used for television broadcasting, there is a high possibility that a moving image is compressed and transmitted immediately before being broadcast from a broadcasting station to a general home. In such a usage method, it is sufficient that the broadcasting station has at least one encoder. Broadcast materials and commercial materials are stored or transmitted without being compressed. Edit the uncompressed material in a predetermined order (unify), then MP
EG2 will be compressed.

However, when the MPEG2 encoder becomes cheaper and lighter due to technological innovation, it is expected that broadcast materials and commercial materials will also be MPEG2 compressed and transmitted or stored. In such an application, bitstreams that have been MPEG2 compressed in advance are edited by a broadcasting station and sent to a general home.

When cutting an MPEG2-compressed bit stream for editing, it is difficult to cut and edit in the middle of the GOP described above. Therefore, the GOP is edited. In the example shown in FIGS. 7A and 7B, there is a point that can be cut every 10 frames.
In the example shown in (c), there are cuttable points every 6 frames.

In the video material used for broadcasting, a number called a time code is added to each frame. When editing video material, the edit point is determined by referring to this time code. The time code is
It consists of four numbers: hour: minute: second: frame.

The NTSC signal used for television broadcasting in Japan and the United States transmits 29.97 images per second. Since this value is approximately 30, the frame number is represented by a value from 0 to 29, but it is partially discontinuous because it deviates from 30 by 0.1%.

Specifically, 0 minutes, 10 minutes, 20 minutes every hour,
Each minute except 30 minutes, 40 minutes, and 50 minutes (values that are well separated in minutes, that is, seconds and frame values are 00 seconds 00
From the start of (frame), the time code is thinned out every two frames (00 frame and 01 frame). A frame of the thinned time code is called a drop frame. In this way, by thinning out 0.1% frames, it is possible to prevent the time lag between the real time and the time code from being accumulated.

As described above, since the number of frames per second of the NTSC signal is not an integer, when the GOP is constructed by dividing the video signal into equal frame intervals as in the conventional method,
The relationship between the GOP head frame and the positive segment is slightly shifted. For example, if the first GOP is started at 00 minutes 00 seconds 00 frames and divided by 10 frames each, 01 minutes 00 seconds 02 frames, 02 minutes 00 seconds 0
Four frames, 03 minutes 00 seconds 06 frames, and 04 minutes 00 seconds 08 frames are the GOP heads, respectively. As described above, even if the compression is started so that the minute and the second are the heads of the GOP first, there is a drawback that the compression gradually shifts.

A broadcasting station edits a video signal according to a predetermined time schedule and sends it to a general home, but the time schedule is often decided in units of minute or second. Therefore, even when a bitstream is edited to create a transmission signal, it must be possible to edit in units of minutes and seconds. In the conventional GOP configuration,
First, even if the compression is started so that the GOP head is a minute and a second, the GOP head is not guaranteed because it gradually shifts. Therefore, it is expected that editing work will be hindered.

The present invention has been made in order to solve the above-mentioned conventional problems, and it is an object of the present invention to provide a moving picture compression apparatus capable of aligning time code minutes and seconds to the boundary of an image aggregate. To aim.

[0021]

A moving picture compression apparatus according to the present invention is a moving picture compression apparatus which divides a moving picture into image aggregates of predetermined frames and compresses the time code of each frame. Time code input means or time code generation means for generating time code,
A drop frame predicting and detecting means for predicting and detecting a drop frame in the time code; and an image aggregate determining means for determining the number of images of the image aggregate based on the results of the drop frame predicting and detecting means. is there.

[0022]

According to the MPEG2 standard, it is possible to change the GOP length in the middle of the bit stream. The presence of a drop frame in the time code is detected by the drop frame prediction / detection means, and if there is a drop frame, the length of GOP is changed to perform moving image compression.

By compressing the moving image in this way, it is possible to create a bit stream in which every minute and every second are GOP heads.

[0024]

1 is a block diagram showing the configuration of a moving picture compression apparatus according to an embodiment of the present invention. The configuration and operation of this embodiment will be described with reference to FIG.

In FIG. 1, 1 is an image signal input terminal and 1
Reference numeral 01 is a time code input terminal, 103 is a drop frame prediction / detection means, 104 is a GOP configuration determination means, 2 is an encoding means, and 3 is a bit stream output terminal.

In the moving picture compression apparatus of this embodiment, the moving picture signal inputted from the picture signal input terminal 1 and the time code inputted from the time code input terminal 101 are compressed in accordance with a predetermined standard, and the bit is compressed. It has a function of outputting from the stream output terminal 3.

The moving image signal input from the image signal input terminal 1 is synchronized with the time code input from the time code input terminal 101, and this time code includes a drop frame. As for the time code, the drop frame prediction / detection means 103 determines whether or not there is discontinuity in the time code in the GOP to be compressed, and based on this result, the GOP configuration determination means 104 determines the GOP. Of the GOP and the I, P, and B pictures to which each frame is to be compressed are determined.
Compressed in long and picture type.

The drop frame prediction, the operation of the detecting means 103, and the operation of the GOP configuration determining means 104 will be described with reference to FIG.

FIG. 2 is a flow chart showing a method for determining the length of the GOP and the configuration. In FIG. 2, A101
The area surrounded by is a flow chart showing the drop frame prediction in FIG.
The area surrounded by 102 is a flow chart showing the judgment by the GOP configuration determining means 104 in FIG. 1 (S1).
(S12) shows each step. The length of the GOP is determined by the flow of the area A101, and in response to the result, the GOP structure is determined by the flow of the area A102 in FIG.

GOP when no drop frame is included
Let N be the length and M be the period of the I or P picture.
It corresponds to N and M in FIG. First, the GOP length N is determined. Therefore, it is determined using the input time code whether or not a drop frame is included in the N frames to be compressed (S1). If it is not included, the GOP length is set to N. To set the GOP length to N,
The initial value of the frame counter n used in the area A102 is set to 0 (S3). If a drop frame is included in step (S1), it is further determined whether or not a drop frame is included in the N-1 frames to be compressed (S2). If a drop frame is included, the GOP length is set to N-2 and the initial value of the frame counter n is set to 2 (S4). If it is not included in the step (S2), the GOP length is set to N-1, and the frame counter n
The initial value of is set to 1 (S5).

FIG. 3 shows the area A by taking the case of N = 6 as an example.
The relationship between the judgment standard and the judgment result in the flow 101 is GO
It is the figure shown by the value of the time code in P boundary.

FIG. 3A shows an example in which no drop frame exists between the frame to be compressed and N frames. In this case, even if the GOP length is N, the drop frame does not exist in the GOP. Therefore, the right-hand frame that comes immediately after the drop frame (see FIG.
In (a), 00: 01: 00: 02 is the GOP head frame. This state is the step (S3) in FIG.
Equivalent to.

FIG. 3B shows an example in which a drop frame exists between the frame to be encoded and N frames. In this case, by setting the GOP length to N-1, it is possible to encode so that the drop frame does not exist in the GOP. At this time, the frame of the right segment that comes immediately after the drop frame (in FIG. 3B,
0: 01: 00: 02) is the first frame of the GOP. This state corresponds to step (S5) in FIG.

FIG. 3C is an example in which a drop frame exists between the frame to be encoded and the N-1 frame. In this case, by setting the GOP length to N-2, it is possible to encode so that the drop frame does not exist in the GOP. At this time, the frame of the correct number (0: 01: 00: 02 in FIG. 3C) immediately after the drop frame is the GOP head frame. This state corresponds to step (S4) in FIG.

As described above, the drop frame predicting / detecting means 103 in FIG. 1 predicts the existence of a drop frame, and as a result, the GOP configuration determining means 104 in FIG. I in the frame,
Assign P and B pictures.

In the area A102 of FIG. 2, the picture type is determined based on the value of the frame counter n (S).
6).

First, by determining the remainder when n is divided by M, it is determined whether or not the B picture should be used. If the remainder obtained by dividing n by M is not equal to (M-1), the B picture is selected (S8). In step (S6), the operator represented by% is a binary operator that calculates the remainder by dividing the left term by the right term. If it is not a B picture, it is compressed as an I picture or a P picture. In the GOP, the first frame that is not a B picture is compressed as an I picture, and the subsequent frames that are not B pictures are P pictures.
It is compressed as a picture (S7, S9, S10).

After determining the picture type in this way, the frame counter n is incremented by 1 (S1).
1). When the frame counter becomes equal to the GOP length N (S12), in order to start the processing of the next GOP,
Transition to the state of step (S1). Otherwise, the process proceeds to step (S6) to move to the process of determining the picture type of the next picture.

According to the picture type determined in this way, compression is performed in the encoder 2 in FIG.
It is output from the bitstream output terminal 3.

FIG. 4 shows G of the moving image encoded by this embodiment and the moving image encoded by the conventional method.
It is a figure showing an example of OP composition.

In the figure, 201 is an example of a time code of an input image and represents hours: minutes: seconds: frames. Reference numeral 211 is an example of GOP division when encoded according to this embodiment, and reference numerals 212, 213, and 214 are GOPs. On the other hand, 221 is an example of GOP division when encoded by the conventional method.
2, 223 and 224 are GOPs, respectively.

In this example, 1 GOP consists of 10 frames (N = 10), and 0: 00: 59: 20 is 20 frames.
It shows the relationship between the GOP start frame and the positive component when the OP starts.

In the first GOP, since there are no drop frames within the N frame period, the GOP is composed of 10 frames in this embodiment as in the conventional method. In the next minute (0: 01: 00: 02 frame), in this embodiment,
In both the conventional examples, the head of the GOP is the exact minute.

In the next GOP, since the drop frame exists in the 0th frame, in this embodiment, the G frame is set in 8 frames.
Configure OP. On the other hand, in the conventional method, since the number of frames in the GOP is constant, the GOP is composed of 10 frames. As a result, in the present embodiment, the next GOP starts from 00: 010: 00.10 frames, whereas in the conventional method, the next GOP starts from 00: 01.00.12 frames. In the method, the GOP start frame number becomes a value that is not divisible by N. Since this deviation is maintained as it is, the next correct minute is 0: 02: 0.
In the 0 second 02 frame, in the present embodiment, the GOP head is the exact minute, whereas in the conventional method, the GOP head and the exact minute do not match. The deviation is further increased in the following correct minutes.

FIG. 6 (a) is a diagram showing which number of frames in the GOP each positive component corresponds to when encoded under the same conditions as in FIG. In the present embodiment, the frame number in the GOP is 0 (that is, the head of the frame) in the straight line, whereas the frame number in the GOP gradually shifts in the conventional method.

In FIG. 4 and FIG. 6A, the case where the number of frames in the GOP is 10 (N = 10) has been described as an example.
According to this method, as long as the number of frames in one minute when no drop frame is included can be divided by N, all other time code Can be aligned with a GOP boundary. In the case of an NTSC image, the GOP length N satisfying the following conditions is sufficient.

(1800% N) = 0%: Binary operator for calculating the remainder by dividing the left term by the right term. Specifically, for 36 N values shown in FIG. 6B, It is possible to match all time code minutes to GOP boundaries.

FIG. 5 is a block diagram showing the structure of a moving picture compression apparatus which is another embodiment of the present invention.

The present embodiment is an example in which the time code is not supplied from the outside but is generated by the time code generating means 102 and is encoded together with the input moving image. In this example, the time code generating means 102 can perform the same GOP length control as in the first embodiment by performing the same processing as the flow represented by the area A101 in FIG. The effect of can be obtained.

In the above embodiment, the image aggregate is M
Assuming "GOP" of PEG standard,
An example of encoding so that the exact second coincides with the GOP boundary has been described. However, assuming that a “sequence” defined by the MPEG2 standard is used as an image aggregate, the sequence boundary coincides with the time code minute and the exact second. It is also possible to encode.

[0051]

As described above, according to the present invention, in a moving image compression apparatus for dividing a moving image into image aggregates of predetermined frames and compressing the image, time code input means for inputting the time code of each frame. (101) or a time code generation means (102) for generating a time code, a drop frame prediction / detection means (103) for predicting and detecting a drop frame in the time code, and a drop frame prediction / detection means (103). ) Image group determining means (1) for determining the number of images of the image group based on the result of
04), it is possible to align the time code minutes and seconds with the image aggregate boundaries, which facilitates editing after creating the bitstream.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.

FIG. 2 is a flowchart showing a method of determining the length and configuration of a GOP according to the present invention.

FIG. 3 is a diagram showing the relationship between the determination standard and the determination result in the flowchart of FIG. 2 when the GOP length N is 6 by the time code value at the GOP boundary.

FIG. 4 is a diagram showing an example of GOP configurations of a moving image encoded according to an embodiment of the present invention and a moving image encoded by a conventional method.

FIG. 5 is a block diagram showing the configuration of another embodiment of the present invention.

FIG. 6 is a diagram showing the order in the GOP of each positive segment when encoded under the same conditions as in FIG. 4;
It is a figure which shows that it is possible to make a tie-code positive part correspond to a GOP boundary with respect to a normal N value.

FIG. 7 is a diagram showing various configuration examples of a GOP.

[Description of Reference Signs] 1 image signal input terminal 2 encoding means 3 bit stream output terminal 101 time code input terminal 102 time code generation means 103 drop frame prediction and detection means 104 GOP configuration determination means

Front page continuation (72) Inventor Toshifumi Sakaguchi 4-36-19 Yoyogi, Shibuya-ku, Tokyo Inside Graphics Communications Laboratories, Inc. (72) Yuzumi Wataya 4-36-19 Yoyogi, Shibuya-ku, Tokyo No. Stock Company Graphics Communications Laboratories

Claims (1)

[Claims] 1. A moving picture compression apparatus for dividing a moving picture into image aggregates of predetermined frames and compressing the same, and generating time code input means (101) for inputting a time code of each frame or a time code. The time code generation means (102), the drop frame prediction and detection means (103) for predicting and detecting the drop frame in the time code, and the image set based on the results of the drop frame prediction and detection means (103). A moving image compression apparatus, comprising: an image aggregate determining means (104) for determining the number of images of a body.