JPH1042295A - Video signal encoding method and video signal encoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To satisfy the frame rate of an MPEG standard and to reduce a practical frame rate. SOLUTION: A copy stream generator 19 for generating a code word bit stream (copy stream) for indicating that it is the same as a previously encoded picture is provided. A system control part 15 supplies the code word bit stream (copy stream) 19a outputted from the copy stream generator 19 through an output data changeover switch 16 for comprising a code multiplexer to a butter 7 without performing inter motion compensation frame encoding to a B picture or a P picture to be the object of frame thinning. By inserting the code word but stream of just copying a frame in front or at the back cyclically, the frame rate lower than a normal frame rate is realized.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video signal encoding method and a video signal encoding device, and more particularly, to a standard frame rate by inserting data designating frame copying into a bit stream. The present invention relates to a video signal encoding method and a video signal encoding device capable of substantially reducing the frame rate while satisfying the following conditions.

[0002]

2. Description of the Related Art MPEG1 (ISO / IEC11172) is known as one of moving picture coding methods for storage media (MPEG: Moving Picture).
Experts Group). FIG. 1 shows the data structure of MPEG1, and FIG. 2 shows the bit stream structure of MPEG1.
7 to FIG. FIGS. 1 to 7 are described in Chapter 6 of "International Standards for Multimedia" (published by Maruzen), edited by Hiroshi Yasuda. These data structure and bit stream structure relate to a standardization plan.

As shown in FIG. 1, MPEG1 has a data structure of six layers of blocks, macroblocks, slices, pictures, GOPs, and sequences.

A block is composed of 8 pixels × 8 lines of pixels having adjacent luminance or color differences. DCT (Discrete Cosine Transform) is executed on a block-by-block basis.

A macroblock is composed of four luminance blocks Y0, Y1, Y2, and Y3 adjacent to each other in the left and right and up and down directions and color difference blocks Cb and Cr at the same position on the image.
It consists of one block. The transmission order is Y0, Y1,
Y2, Y3, Cb, and Cr. It is determined in macroblock units what to use for prediction data (reference image data for obtaining a difference: created by forward prediction, backward prediction, bidirectional prediction, etc.), and whether or not to send a difference. If the coded block is the same as the predicted macroblock, no macroblock layer data is sent. It is said to skip this. When a macroblock is skipped in succession, the next non-skipped macroblock has the number of previous skipped macroblocks.

[0006] A slice is composed of one or a plurality of macroblocks connected in the order of image operations. At the beginning of the slice, the difference between the motion vector and the DC component in the image is reset. The first macroblock has data indicating the position in the image, and can be restored even if an error occurs. The length of the slice and the starting position are arbitrary. The first and last macroblocks in a slice must be non-skipping macroblocks.

[0007] A picture (an image for each image) is composed of at least one or a plurality of slices. The picture is
According to the encoding method, the picture types are classified into three types, I, P, and B.

[0008] I picture (Intra-coded p
image: intra-coded image) uses only closed information in one image when it is encoded.
In other words, when decoding, an image can be reconstructed using only the information of the I picture itself.

A P-picture (Predictive-co
The ded picture (forward predicted coded image) uses an I-picture or a P-picture which is located temporally earlier in the input and has already been decoded, as a predicted image (a reference image for taking a difference). In practice, it is possible to select, on a macroblock-by-macroblock basis, whether to encode the difference from the motion-compensated predicted image or to encode without taking the difference (intra encoding).

[0010] B picture (Bidirectional p
(reactive-coded picture) is a previously decoded I picture or P picture which is temporally located earlier as a predicted picture,
It uses three types of interpolated images, which are located later in time and have already been decoded from I-pictures or P-pictures, and both. The most efficient of the three types of difference coding and intra coding after motion compensation can be selected in macroblock units.

[0011] A GOP (group of pictures) is composed of one or more I pictures and zero or more non-I pictures. The input order to the encoder is I1, B2,
B3, P4 | B5, B6, I7, B8, B9, I10,
B11, B12, P13, B14, B15, P16 | B
17, B18, I19, B20, B21, and P21, the outputs of the encoder are I1, P4, B2, B3 | I7,
B5, B6, I10, B8, B9, P13, B11, B
12, P16, B14, B15 | I19, 17, B1
8, P21, B20, and B21. Where I, P,
B represents a picture type, numbers represent an input order to the encoder, and | represents a GOP break. In order to encode or decode a B picture, it is necessary that an I picture or a P picture which is a temporally backward prediction picture is previously coded. A change (order change) is made. I picture spacing and P
The intervals between pictures are free and may vary within a GOP.

A sequence (video sequence) is composed of one or a plurality of GOPs having the same image size, image rate, and the like.

As shown in FIG. 2, the syntax of a bit stream is defined for each layer. For example, in the sequence (video sequence) layer, a synchronization code SSC indicating the beginning of the sequence layer, the number of horizontal pixels HS of the image, the number of vertical lines VS of the image, an index PAR indicating the aspect ratio of the pixel interval, the display rate of the image And various contents, such as the index PR, are specified. The meanings of the abbreviations in FIG. 2 are shown in FIGS.

FIG. 8 is a block diagram of a conventional MPEG1 encoder. Conventional MPEG1 encoder 101
Is a motion prediction unit 102, a subtractor 103, a discrete cosine transformer (DCT) 104, a quantizer 105, a variable length encoder 106, a buffer 107, and an inverse quantizer 1
08, an inverse discrete cosine transform (inverse DCT) 109,
An adder 110, a delay circuit 111, a first frame buffer 112, a second frame buffer 113, an average calculation circuit 114, a system control unit 115, and an output configuring a multiplexing unit for encoded data. A data changeover switch 116, a reference image selection switch 117 for selecting a reference image for obtaining a difference, and each frame buffer 112, 1
And a local decoded image input changeover switch 118 for changing over the input to the input terminal 13.

An input image is supplied to a motion prediction unit 102. The input image is a frame image including the digital luminance signal Y and the digital color difference signals Cb and Cr. The motion prediction unit 102 can temporarily store a plurality of frame images and perform rearrangement in units of frames. The motion prediction unit 102 performs motion prediction and outputs a motion vector 102a. The motion prediction is to predict how close a macroblock to be predicted is from a macroblock at the same position in a reference frame and closest to the macroblock. The criterion of closeness differs depending on the algorithm, but is usually a value that is considered to minimize the amount of generated bits during compression.

The encoding procedure will be described based on the frame structure shown in FIG. In FIG. 9, 1 in the frame
The letter alphabet represents the picture type. I
Is an I picture, P is a P picture, and B is a B picture. The number following the picture type indicates the encoding order. As described above, the input order and the encoding order are different.

Since the frame to be encoded first has no frame that can be used for prediction, it is assigned to an I picture, and encoding is performed only within the frame. I_
At the time of encoding 0, the system control unit 115 controls the reference image selection switch 117 to a position where data 0 is supplied. As a result, data of 0 is supplied to the reference image data input terminal 103a which takes the difference of the subtractor 103.
Therefore, the pixel (pixel) data 102b of the I picture output from the motion prediction unit 102 is
Is not affected at all, and the pixel data 10 of the I picture
2b is supplied to the discrete cosine transform unit (DCT) 104 as it is. The discrete cosine transform unit (DCT) 104
The pixel data 102b is orthogonally transformed in units of 8 × 8 pixels, and a transform coefficient 104a is output. The transform coefficient 104a is quantized by the quantizer 105. The quantized transform coefficient 105a is converted into a variable length code
is converted to a. Variable length code 106a related to transform coefficient
Is input to the buffer 107 via the output data changeover switch 116, and supplied to, for example, a host computer or the like via the buffer 107.

The system control unit 115 generates encoded data 115a in a layer above the macroblock layer, and switches an output data changeover switch 116 from the output side of the variable length encoder 106 to the output side of the system control unit 115. Then, the coded data 115a related to the sequence layer, the GOP layer, the picture layer, and the slice layer is supplied to the buffer 107.

The transform coefficient 1 output from the quantizer 105
05a is supplied to the inverse quantizer 108 and inversely quantized. The transform coefficient 108 a decoded by the inverse quantization is subjected to inverse orthogonal transform by the inverse discrete cosine transformer 109. The data 109a generated by the inverse orthogonal transform is added to the adder 11
0 is supplied to one input terminal. In the case of encoding an I picture, data of 0 is supplied to the other input terminal of the adder 110 via the delay circuit 111, so that the data 109a output from the inverse discrete cosine transformer 109
Is output as it is without being affected by the adder 110.

At the time of encoding I_0, the system control unit 115 controls the local decoded image input changeover switch 118 to, for example, the first frame buffer 112 side, so that the locally decoded image of the I_0 frame output from the adder 110 is , Are stored in the first frame buffer 112.

Here, the decoded image data is generated by decoding the image data compressed by the discrete cosine transformer 104 and the quantizer 105 and stored in each of the frame buffers 112 and 113. This is because it is used for the next inter-frame prediction.

Next, a case will be described in which B_1 shown in FIG. 9 is encoded using only backward prediction from I_0. In this case, the system control unit 115 switches the reference image selection switch 117 to the first frame buffer 112 side. Further, the system control unit 115 sets the local decoded image input changeover switch 118 to the off position (a position not connected to any of the frame buffers 112 and 113).
Switch to.

The system control unit 115 includes the motion prediction unit 10
1 to output the pixel data of the macroblock relating to the frame B_1 and the first frame memory 1
12, pixel data of a rectangular area shifted from the same macroblock position as that output by the motion prediction unit 101 by the amount specified by the motion vector 102a is read. The pixel data 112a read from the first frame memory 112 according to the motion vector 102a is supplied to the reference image input terminal 103a of the subtractor 103 via the reference image selection switch 117. Therefore, the subtractor 103 outputs difference data between the pixel data of the B_1 frame image and the pixel data obtained by performing motion compensation on the previously encoded I picture. This difference data is orthogonally transformed by the discrete cosine transformer 104 and quantized by the quantizer 105.
The data is converted into a variable length code by the variable length encoder 106 and stored in the buffer 107 via the output data switch 116.

Since the B picture is not used as a reference image for inter-frame prediction, it is not necessary to generate a locally decoded image by the inverse quantizer 108, the inverse discrete cosine transformer 109, and the like. Therefore, the system control unit 115 sets the inverse quantizer 10
8. Control to prevent the operation of generating a locally decoded image by the inverse discrete cosine transformer 109 or the like.

Next, B_2 shown in FIG. 9 is encoded using only backward prediction from I_0. This encoding operation is the same as the above-described encoding of B_1.

Next, P_3 shown in FIG. 9 is encoded using only forward prediction from I_0. in this case,
The system control unit 115 includes a reference image selection switch 117
To the first frame buffer 112 side. Further, the system control unit 115 switches the local decoded image input changeover switch 118 to the second frame buffer 113 side.

The system control unit 115 controls the motion prediction unit 10
1 to output the pixel data of the macroblock relating to the frame P_3 and the first frame memory 1
12, pixel data of a rectangular area shifted from the same macroblock position as that output by the motion prediction unit 101 by the amount specified by the motion vector 102a is read. The pixel data 112a read from the first frame memory 112 according to the motion vector 102a is supplied to the reference image input terminal 103a of the subtractor 103 via the reference image selection switch 117. Therefore, the subtractor 103 outputs difference data between the pixel data of the P_3 frame image and the pixel data obtained by performing motion compensation on the previously encoded I picture. This difference data is orthogonally transformed by the discrete cosine transformer 104 and quantized by the quantizer 105.
The data is converted into a variable length code by the variable length encoder 106 and stored in the buffer 107 via the output data switch 116.

Since the frame data of the P picture is used for inter-frame prediction, local decoding is performed. The quantized transform coefficient 105 a is inversely quantized by an inverse quantizer 108, inversely orthogonally transformed by an inverse discrete cosine transformer 109, and data 109 a decoded by the inverse orthogonal transform is added to an adder 1.
10 to one input terminal. The other input terminal of the adder 110 is supplied with the pixel data 112a read out from the first frame memory via the delay circuit 111 in accordance with the motion vector 102a. Delay circuit 1
The delay time of the discrete cosine transformer 104, the quantizer 105, the inverse quantizer 108, and the inverse discrete cosine transformer 1
09 is set to the processing time of the encoding / decoding loop formed by the setting of the encoding loop. The locally decoded data of the difference data and the reference image data are added by the adder 110 to generate locally decoded image data. The local decoded image data output from the adder 110 is stored via the local decoded image input changeover switch 118 to the second frame buffer 113 in which no predicted image is stored. Thus, the second frame buffer 113 stores the currently decoded frame data of the locally decoded image of the P picture.

Next, B_4 shown in FIG. 9 is encoded. For the encoding of B_4, three types of prediction modes can be used: forward prediction from I_0, backward prediction from P_3, and bidirectional prediction from the average value of I_0 and P_3. Which prediction mode to use can be selected for each macroblock. The system control unit 115 controls the reference image selection switch 117 based on the prediction mode set for each macro block. At present, the first
, The locally decoded image data of I_0 is stored in the frame buffer 112, and the locally decoded image data of P_3 is stored in the second frame buffer 113. Therefore, in the forward prediction mode from I_0, the reference image selection switch 117 is controlled so as to select the first frame buffer 112 side. In the backward prediction mode from P_3, the reference image selection switch 117 is controlled so as to select the second frame buffer 113 side. In the bidirectional prediction mode, the reference image selection switch 117 is controlled so as to select the output of the average value calculation circuit 114. Here, the average value calculation circuit 114 is connected to the first frame buffer 112
And an average value of the output 112a of the second frame buffer 113 and the output 113a of the second frame buffer 113. The encoding operation is the same as in the case of B_1 described above. B_
5, the same operation as B_4 is performed.

The input image is encoded by repeating such processing. Here, pixel data of the input image output from the motion prediction unit 102,
After being encoded, the frame memory 1 is locally decoded.
The frame data stored in 12 and 113 do not have the same value. The main cause is that in order to reduce the data amount of a bit stream generated by encoding at the time of quantization by the quantizer 105, the discrete cosine transformer 104 is determined by a weighting value called a quantization scale.
This is because the orthogonal transform output is divided and the decimal part is rounded. Therefore, when the data amount of the generated bit stream is limited to a small value, even if the input image is a still image, each frame buffer 11
In order for the locally decoded image data stored in 2,113 to converge to a value close to the input image data, several frames must be passed.

[0031]

As described above, M
PEG1 is based on a method in which motion compensation is performed to reduce redundancy in the time axis direction to obtain a difference between images, and then discrete cosine transform and variable length coding are performed to reduce redundancy in the spatial axis direction. , Bit stream definition and decoding logic.

Here, in MPEG1, as shown in (a3) image rate in FIG. 5, the number of pictures (frames) per second is 23.97 as the frame rate.
6, 24, 25, 29.97, 30, 50, 59.4,
Eight 60 types are defined. Here, the number of frames per second 59.94 is the field rate of the NTSC system,
50 is the PAL field rate, 29.97 is N
The TSC frame rate, 25 is the PAL frame rate, and 23.976 is the 4/5 frame rate of the NTSC frame rate.

On the other hand, depending on the application such as a video conference system and a surveillance camera device, the frame rate may be reduced.
There is a demand to reduce the amount of data. In many cases, it is known that reducing the frame rate reduces the data amount. However, according to the MPEG1 standard, the frame rate cannot be reduced to 23.976 frames / second or less.

The present invention has been made to solve such a problem, and provides a technique capable of reducing a substantial frame rate and a data amount while satisfying a frame rate of the MPEG standard. The purpose is to:

[0035]

According to the present invention, there is provided a video signal encoding method comprising: generating a frame signal indicating that a frame is the same image signal; and inserting the frame signal into a bit stream. Reduce the substantial frame rate.

The video signal encoding apparatus according to the present invention includes a frame copying means for generating a frame signal indicating that the frames are the same image signal and inserting the frame signal into a bit stream, thereby substantially reducing the frame rate. Let it.

[0037]

Embodiments of the present invention will be described below with reference to the accompanying drawings. FIG. 10 is a block diagram of a video signal encoding apparatus according to the MPEG1 standard according to the present invention. The video signal encoding apparatus 1 conforming to the MPEG1 standard includes a motion estimator 2, a subtractor 3, a discrete cosine transformer (DCT) 4, a quantizer 5, a variable length encoder 6, and a buffer 7. , An inverse quantizer 8, an inverse discrete cosine transform (inverse DCT) 9, an adder 10, a delay circuit 11, a first frame buffer 12, a second frame buffer 13, and an average value operation A circuit 14, a system control unit 15, an output data changeover switch 16 constituting a multiplexing unit for coded data, a reference image selection switch 17 for selecting a reference image for obtaining a difference, and a switch to each of the frame buffers 112 and 113. It comprises a local decoded image input changeover switch 18 for switching the input, and a copy stream generator 19 constituting a frame copying means. The difference from the conventional MPEG1 encoder shown in FIG. 8 is that a copy stream generator 19 is provided.

The input image is supplied to the motion prediction section 2. The input image is a frame image including the digital luminance signal Y and the digital color difference signals Cb and Cr. Motion prediction unit 2
Temporarily stores a plurality of frame images and rearranges them in frame units. The motion prediction unit 2 performs a motion prediction and outputs a motion vector 2a.

A normal encoding procedure without using a copy frame will be described with reference to the frame configuration shown in FIG. The first frame to be encoded is an I picture, and encoding is performed only within the frame. At the time of encoding I_0, the system control unit 15 controls the reference image selection switch 17 to a position where data 0 is supplied. Thereby, the reference image data input terminal 3a for obtaining the difference of the subtractor 3
Is supplied with 0 data. Therefore, the pixel (pixel) data 2b of the I picture output from the motion prediction unit 2 is not affected by the subtractor 3 at all, and the pixel data 2b of the I picture is directly used as the discrete cosine transform unit (DC
T) 4. Discrete cosine transform unit (DCT) 4
Converts the pixel data 2b orthogonally in units of 8 × 8 pixels and outputs a conversion coefficient 4a. The transform coefficient 4a is quantized by the quantizer 5. The quantized transform coefficient 5a is converted by the variable length encoder 6 into a variable length code 6a. The variable length code 6a relating to the transform coefficient is input to the buffer 7 via the output data changeover switch 6, and is supplied to the host computer or the like via the buffer 7.

The system control unit 5 generates the encoded data 5a of the layer above the macroblock layer, and switches the output data changeover switch 16 from the output side of the variable length encoder 6 to the output side of the system control unit 15. And the sequence layer,
The coded data 15a relating to the GOP layer, the picture layer, and the slice layer is supplied to the buffer 7.

The transform coefficient 5a output from the quantizer 5
Is supplied to an inverse quantizer 8 and inversely quantized. The transform coefficient 8a decoded by the inverse quantization is subjected to inverse orthogonal transform by the inverse discrete cosine transformer 9. The data 9a generated by the inverse orthogonal transform is supplied to one input terminal of the adder 10. In the case of encoding an I-picture, data of 0 is supplied to the other input terminal of the adder 10 via the delay circuit 11, so that the data 9a output from the inverse discrete cosine transformer 9 is added. It is output as it is without any action by the device 10.

At the time of encoding I_0, the system control unit 15 controls the local decoded image input changeover switch 18 to, for example, the first frame buffer 12, so that the locally decoded image of the I_0 frame output from the adder 10 is , Are stored in the first frame buffer 12.

Here, the image data compressed by the discrete cosine transformer 4 and the quantizer 5 is decoded to generate a decoded image, and the decoded image data is stored in each of the frame buffers 12 and 13 because the encoded frame data is This is because it is used for the next inter-frame prediction.

Next, a case will be described where B_1 shown in FIG. 9 is encoded using only backward prediction from I_0. In this case, the system control unit 15 switches the reference image selection switch 17 to the first frame buffer 12 side. Further, the system control unit 15 switches the local decoded image input changeover switch 18 to an off position (a position not connected to any of the frame buffers 12 and 13).

The system control unit 15 causes the motion prediction unit 1 to output the pixel data of the macroblock of the frame B_1, and outputs the pixel data from the first frame memory 12
The pixel data of a rectangular area shifted from the same macroblock position as that output by the motion prediction unit 1 by the amount specified by the motion vector 2a is read.
The pixel data 12a read from the first frame memory 12 according to the motion vector 2a is supplied to the reference image input terminal 3a of the subtractor 3 via the reference image selection switch 17.
Supplied to Accordingly, the pixel data of the B_1 frame image and the previously encoded I
The difference data from the pixel data obtained by performing motion compensation on the picture is output. This differential data is orthogonally transformed by the discrete cosine transformer 4, quantized by the quantizer 5, converted into a variable length code by the variable length encoder 6, and stored in the buffer 7 via the output data switch 16. Is done.

Since the B picture is not used as a reference image for inter-frame prediction, it is not necessary to generate a locally decoded image by the inverse quantizer 8, the inverse discrete cosine transformer 9, and the like. Therefore, the system control unit 15 controls so as not to perform the operation of generating the locally decoded image by the inverse quantizer 8, the inverse discrete cosine transformer 9, and the like.

Next, B_2 shown in FIG. 9 is encoded using only backward prediction from I_0. This encoding operation is the same as the above-described encoding of B_1.

Next, P_3 shown in FIG. 9 is encoded using only forward prediction from I_0. in this case,
The system control unit 15 switches the reference image selection switch 17 to the first frame buffer 12 side. Further, the system control unit 15 switches the local decoded image input changeover switch 18 to the second frame buffer 13 side.

The system control unit 15 causes the motion prediction unit 1 to output the pixel data of the macroblock relating to the frame P_3, and from the first frame memory 2, the macroblock position output by the motion prediction unit 1 Pixel data of a rectangular area shifted from the same macroblock position by the amount specified by the motion vector 2a is read. The pixel data 12a read from the first frame memory 12 according to the motion vector 2a is supplied to the reference image input terminal 3a of the subtractor 3 via the reference image selection switch 17. Therefore, from the subtractor 3, P_
The difference data between the pixel data of the frame image No. 3 and the pixel data obtained by performing motion compensation on the previously encoded I picture is output. This differential data is orthogonally transformed by the discrete cosine transformer 4, quantized by the quantizer 5, converted into a variable length code by the variable length encoder 6, and stored in the buffer 7 via the output data switch 16. Is done.

Since the frame data of the P picture is used for inter-frame prediction, local decoding is performed. The quantized transform coefficient 5a is inversely quantized by the inverse quantizer 8, inversely orthogonally transformed by the inverse discrete cosine transformer 9, and data 9a decoded by the inverse orthogonal transform is supplied to one input terminal of the adder 10. Supplied to The other input terminal of the adder 10 is supplied with the pixel data 12a read from the first frame memory via the delay circuit 11 in accordance with the motion vector 2a. The delay time of the delay circuit 10 is set to the processing time of the encoding / decoding loop formed by the discrete cosine transformer 4, the quantizer 5, the inverse quantizer 8, and the inverse discrete cosine transformer 9. I have. The locally decoded data of the difference data and the reference image data are added by the adder 10 to generate locally decoded image data. The local decoded image data output from the adder 10 is stored via the local decoded image input changeover switch 18 to the second frame buffer 13 where no predicted image is stored. This allows
The second frame buffer 13 stores the currently decoded frame data of the locally decoded image of the P picture.

Next, B_4 shown in FIG. 9 is encoded. For the encoding of B_4, three types of prediction modes can be used: forward prediction from I_0, backward prediction from P_3, and bidirectional prediction from the average value of I_0 and P_3. Which prediction mode to use can be selected for each macroblock. The system control unit 15 controls the reference image selection switch 17 based on the prediction mode set for each macro block. At the present time, the first frame buffer 12 stores the locally decoded image data of I_0, and the second frame buffer 13 stores the locally decoded image data of P_3. Therefore, I
In the forward prediction mode starting from _0, the reference image selection switch 17 is controlled so as to select the first frame buffer 12 side. In the backward prediction mode from P_3, the reference image selection switch 17 is controlled so as to select the second frame buffer 13 side. In the bidirectional prediction mode, the reference image selection switch 17 is controlled so as to select the output of the average value calculation circuit 14. Here, the average value calculation circuit 14
It is configured to output the average value of the output 12a of the first frame buffer 12 and the output 13a of the second frame buffer 13. The encoding operation is the same as in the case of B_1 described above. The same operation as B_4 is performed for B_5. By repeating such processing, the input image is encoded.

Next, copying of a frame will be described.
The video signal encoding device 1 shown in FIG. 10 is different from the conventional MPEG1 encoder shown in FIG. 8 in that a copy stream generator 19 is provided. When the copy stream generator 19 is supplied with the copy stream output request 15b from the system control unit 15, the copy stream generator 19 outputs a codeword bit stream 19a related to copying indicating that the frame is the same image as the previously encoded frame. It is configured to do. When generating the copy stream output request 15b, the system control unit 15 controls the output data changeover switch 16 to switch to the codeword bit stream 19a side. Therefore, the code word bit stream 19a output from the copy stream generator 19
Is supplied to the buffer 7 via the output data changeover switch 16 and is output to another device (not shown) or a communication line via the buffer 7.

The function of the copy stream generator may be provided in the system controller 15 without providing the copy stream generator 19 as independent hardware. When the function of the copy stream generator is provided in the system control unit 15, the system control unit 15 includes a code word bit for copying a frame in encoded data 15a for a sequence layer, a GOP layer, a picture layer, and a slice layer. The stream 19a is output.

A case where all the B pictures are copied in the backward prediction mode in the frame configuration shown in FIG. 9 will be described. FIG. 11 is an explanatory diagram showing an example of a frame configuration including a copy frame. In FIG. 11, the arrow indicated by a thick line is a portion where copying is used. When this copy frame is designated, the pictures (frames) of B_1 and B_2 are the same as the picture (frame) of I_0. Also, B
_4 and B_5 have the same picture as P_3,
7, B_8 is P_6 and B_13, B_14 are P_12.
Will be the same as

As a result, when the encoded image data including the code word bit stream 19a relating to the copying is decoded, as shown in FIG. 12, the picture of I_0 continues for three frames, and then the picture of P_3 becomes three frames. ,
The P_6 picture has three frames, the P_9 picture has three frames, and the P_12 picture has three frames. Therefore, when the input image is in the NTSC format, the frame rate is 2 according to the MPEG1 standard.
Although 9.97 frames / sec will be used, the effective frame rate can be 1/3 of 29.97 frames / sec at 9.99 frames / sec.

FIG. 13 is a list showing the contents of a code word bit stream relating to copying when the copying destination is a B picture. When copying a previously coded picture without normal coding of the B picture, the copy stream generator 19 outputs a codeword bit stream 19a for copying described below.

In the data structure of MPEG1, as shown in FIG. 2, the PSC, TR, PCT, VD
(BF), FPFV, FF, FPBV, BF, EBP,
It is necessary to send BA etc.

Here, PSC (picture station)
(rt code) is a synchronization code at the beginning of the picture layer, and its bit length is 32 bits. TR (tem
“polar reference” is a value indicating the display order, and its bit length is 10 bits. This TR is G
Reset at the beginning of OP. If TR exceeds 1024, the remainder value at 1024 is used. PCT (pictu
“re codingtype” is a value indicating the encoding mode (picture type) of the image. In the encoding mode of this image, a 3-bit code is defined as shown in FIG. VD (BF) is the amount of data to be stored in the bit stream buffer until decoding of the picture starts, and its bit length is 16 bits. F
PFV (full pel forward vector)
(or) is a code indicating whether the accuracy of the motion vector is a pixel unit or a half pixel unit when a B or P picture exists, and its bit length is 1 bit. FF (forward f) is a code indicating a search range of a forward motion vector, and its bit length is 3 bits. FPBV (full
l pel background vector) is a code indicating whether the accuracy of a motion vector is a pixel unit or a half pixel unit when a B picture exists, and its bit length is 1 bit.
BF (backward f) is a code indicating a search range of a backward motion vector, and its bit length is 3 bits. EBP (extra bitpictur)
e) is a flag indicating the presence or absence of an extra information picture, and its bit length is 1 bit × n. BA (byt
(e alingn) is a dummy bit for byte alignment.

In the data structure of MPEG1, SSC, QS, EBS, etc. need to be transmitted in the picture layer as shown in FIG.
It is necessary to send AI, MBTYPE, MHB, MVB, MBESC, etc.

Here, the SSC (slice start)
(code) is a synchronization code indicating the beginning of the slice layer, and its bit length is 32 bits. QS (qua
“nize scale” is data that gives a quantization width used for the slice, and its bit length is 5 bits. EBS (extra bit slice) is a flag indicating the presence or absence of an extra information slice, and its bit length is 1 bit × n.

MBAI (macroblock add)
(ress increment) is a variable length code indicating the number of skip MBs before the MB + 1, and the number of bits is 1 to 11 bits. MBTYPE (macro
(block type) is a variable length code indicating the encoding mode of the MB, and its bit number is 1 to 8 bits. MHB (motion horizontal b)
ackward) is present when the MB type is backward and bi-directional prediction, and calculates the difference between the horizontal component of the backward motion vector of the MB and the previous vector as backward.
coded in a VLC table represented by f
Its bit length is 1 to 14 bits. MVB (mot
ion vertical back) is M
The type B exists at the time of backward and bidirectional prediction, and the difference between the vertical component of the backward motion vector of the MB and the previous vector is represented by VL represented by backward f.
C, and the bit length is 1 to 1.
4 bits. MBESC (macroblock
(escape) is a code corresponding to 33 skipped macroblocks, and its bit length is 11 bits.

Therefore, the copy stream generator 19
Is configured to transmit the following bit stream 19a in the picture layer, as shown in FIG. First, a 32-bit PSC (picture start)
code) 000000000000000000000
000010000000 is sent out. Next, a 10-bit TR (temporal reference) is transmitted. This TR is not the encoding order but the number of the input order of the video source in the GOP starting from 0. Next, a 3-bit PCT (pict
A code indicating a B-picture is transmitted according to “ure coding type” 011. Next, the 16-bit V
D (vdv delay) is transmitted. Next, FPF
The motion vector information is sequentially transmitted by V, FF, FPBV, and BF. Next, 1-bit EBP (extra
bit picture). Then, at the end of the picture layer, for example, a 2-byte code 00 for byte alignment for transmitting a slice start code for the next slice layer is transmitted.

Then, the copy stream generator 19
As shown in FIG. 13, the following bit stream 19a is transmitted in the picture layer. First, a 32-bit SSC (slice start c)
mode) 0000000000000000000000000
0010000000001 is sent out. Next, data QS (quantize sc giving a quantization width of 5 bits)
ale). Next, a flag EBS (extrabit) indicating the presence or absence of a 1-bit extra information slice
slice).

Further, the copy stream generator 19
As shown in FIG. 13, in the macroblock layer, first, an MBAI (macroblock address) is used.
increment). Next, data of the first macro block is transmitted. First, MBTYPE
For example, a 3-bit code of 010 indicating a B picture is transmitted as (macroblock type). Next, the fact that the value of the motion vector is (0, 0) is transmitted by MHB and MVB, and the fact that the quantized discrete cosine transform coefficients are all 0 is transmitted. Then, a skip macro block is designated from the second macro block onward to the macro block immediately before the last macro block. Here, the 11-bit MBESC
(Macroblock escape) 000000
It is designated by sending 01000.
Since the MBESC corresponds to 33 skipped macroblocks, the required number of MBESCs are repeatedly transmitted. M
The number of BESCs is (total number of macroblocks−
2) / 33. Next, the last macroblock is designated by MBAI. MBAI (mac
roblock address increment
The value of t) is (total number of macroblocks-1)% 3
3. Here,% 33 indicates that the remainder obtained by dividing by 33 is obtained. Then, also for the last macroblock, the value of the motion vector is (0,
0), and that the quantized discrete cosine transform coefficients are all zero. Then, a byte-aligned bit (dummy bit) is transmitted for the next start code.

The case where copying is performed on a B picture has been described above, but the same can be performed on a P picture. FIG. 14 shows a list of code word bit streams for copying when the copy destination is a P picture.

Even if a B picture is copied,
Even if a picture is copied, the bit amount of the bit stream for copying per picture is the same.
Assuming decoding by software, it is desirable to set a B-picture from which decoding can be omitted as a copy destination. This is because, when software decoding is performed, if the processing cannot be completed within a certain period of time, data is read and discarded without decoding the B picture. In the case of a P picture, the decoded frame is used to decode another frame, so that the data cannot be discarded.

The data amount of the bit stream for copying is about 24 for a frame of 352 pixels wide and 240 pixels high.
0 bits. This value is a bit stream rate of 1.152 Mbit / bit used in a video CD or the like.
Less than 0.021 percent when compared to seconds. In the example shown in FIG. 11, since about 20 frames of the NTSC image data of about 30 frames per second are replaced with the bit stream related to copying, the data amount of the bit stream related to copying about 20 frames is as follows. That is about 0.42 percent of the bit rate of 1.152 Mbit / s.

As an example when applied to other fields, 6
Assume encoding at 5,536 kbit / s. The frame size at this time is 160 pixels horizontally and 1 pixel vertically.
Assuming 12 pixels, the data amount of the bit stream for copying is about 160 bits. 2 of total bit amount
The ratio of the data amount of the bit stream relating to the copying of 0 frames is 4.9%.

As described above, by inserting copy frames periodically, it is possible to realize a frame rate smaller than the frame rate specified by MPEG1.

A copy frame may be inserted as needed. When transmitting a bit stream obtained by performing digital image compression using communication in real time, the line may be congested and transfer at a desired bit rate may not be possible. In such a case, the amount of generated bits can be suppressed by using a copy frame, and as a result, continuity of data can be maintained.

In recent years, the processing capability of the microprocessor has been dramatically increased, and the decoding process of the MPEG1 bit stream can be performed by software. However, during the decoding process by the software, the processing capability of the microprocessor is used for the decoding process, which hinders other operations. Therefore, by reducing the frame rate by copying, it is not necessary to perform inverse quantization and inverse discrete cosine transform (inverse DCT) at the time of decoding processing (decompression processing of compressed image data). Can be reduced. Also, by explicitly indicating that this frame is a copy frame in the bit stream, there is little need to process the frame. It should be noted that the copy frame inserted in the bit stream can be easily determined to be a copy frame from its data structure, so that almost no processing is performed on the copy frame.

Next, another configuration example of the video signal encoding apparatus according to the present invention will be described with reference to FIG. A video signal encoding device 21 shown in FIG. 15 includes a quantized data changeover switch 22 and a motion vector data changeover switch 2.
3 constitute frame copying means. Note that the output data changeover switch 24 switches between selecting the output 6a of the variable length encoder 6 and selecting the output 15a of the system control unit 25.

The quantized data changeover switch 22 is interposed between the input side of the variable length encoder 6 and the quantizer 5. The quantized data switch 22 determines whether the input data to the variable length encoder 6 is the output (quantized transform coefficient) 5a of the quantizer 5 or 0 data.
The switching is performed under the control of the system control unit 25.

Motion vector data changeover switch 23
Is configured to switch between the motion vector supplied to each unit and the motion vector 2a output from the motion prediction unit 2 or 0 under the control of the system control unit 25.

When performing a normal encoding operation, the system control unit 25 switches the quantized data so that the output (quantized transform coefficient) 5 a of the quantizer 5 is supplied to the variable length encoder 6. The switch 22 is switched, and the motion vector data switch 23 is switched so that the motion vector 2a output from the motion prediction unit 2 is supplied to each unit.

When generating a copy bit stream by copying a frame, the system control unit 25 switches the quantized data switch 22 so that the data input to the variable length encoder 6 becomes 0, and , The motion vector data switch 23 is switched so that the motion vector becomes zero.

Thus, the motion vector is forcibly set to 0,
By setting the quantization output to 0, the coding block becomes the same as the prediction block, and the macroblock layer data can be skipped without sending any data. Therefore,
The amount of coded data to be transmitted is reduced, and the substantial frame rate can be reduced.

[0078]

As described above, the video signal encoding method and the video signal encoding apparatus according to the present invention provide a code for indicating that the frame to be decimated is the same image as the previously encoded frame. Since the word bit stream is inserted, the substantial frame rate can be reduced while satisfying the standard frame rate condition.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram showing a data structure of MPEG1.

FIG. 2 is an explanatory diagram showing a bit stream structure of MPEG1.

FIG. 3 is an explanatory diagram showing a bit stream structure of MPEG1.

FIG. 4 is an explanatory diagram showing a bit stream structure of MPEG1.

FIG. 5 is an explanatory diagram showing a bit stream structure of MPEG1.

FIG. 6 is an explanatory diagram showing a bit stream structure of MPEG1.

FIG. 7 is an explanatory diagram showing a bit stream structure of MPEG1.

FIG. 8 is a block diagram of a conventional MPEG1 encoder.

FIG. 9 is an explanatory diagram showing an example of an encoding procedure.

FIG. 10 is a block diagram of a video signal encoding apparatus according to the present invention.

FIG. 11 is an explanatory diagram illustrating an example of a frame configuration including a copy frame.

FIG. 12 is an explanatory diagram showing how an apparent frame looks when a copy frame is used.

FIG. 13 is a list showing the contents of a codeword bit stream relating to copying when a copy destination is a B picture.

FIG. 14 is a list showing the contents of a codeword bit stream relating to copying when a copy destination is a P picture.

FIG. 15 is a block diagram of another video signal encoding device according to the present invention.

[Explanation of symbols]

1, 21 video signal encoding device, 2 motion prediction unit, 4
Discrete Cosine Transformer, 5 Quantizer, 6 Variable Length Encoder, 7 Buffer, 8 Inverse Quantizer, 9 Inverse Discrete Cosine Transformer, 15, 25 System Control Unit, 16, 24
Output data switch, 19 copy stream generator, 22 quantized data switch, 23
Motion vector data switch

Claims (6)

[Claims] 1. A video signal encoding method for generating a digitally encoded signal encoded by inter-frame prediction, comprising: generating a frame signal indicating that frames are the same image signal and inserting the frame signal into a bit stream. Characteristic video signal encoding method. 2. A video signal encoding apparatus for generating a digitally encoded signal encoded by inter-frame prediction, wherein a frame signal indicating that a frame is the same image signal is generated and inserted into a bit stream. A video signal encoding device comprising means. 3. The video signal encoding apparatus according to claim 2, wherein said frame copying means comprises a copy stream generator for generating a preset code word bit stream. 4. The video signal encoding apparatus according to claim 2, wherein said frame copying means comprises switching means for forcibly setting a motion vector and a quantized output of a quantizer to zero. 5. The video signal encoding apparatus according to claim 2, wherein said digitally encoded signal is a signal of an MPEG standard. 6. The video signal encoding apparatus according to claim 5, wherein the frame copied by said frame copying means is a B picture.