JPH07131793A - Video signal high efficiency coding device - Google Patents

Abstract

PURPOSE:To improve the coding efficiency. CONSTITUTION:Picture data from a small block processing section 31 are given to an orthogonal transformation section 16 via a switch 33 and given to the orthogonal transformation section 16 via an inter-frame prediction section 32 and the switch 33. A comparison arithmetic section 36 compares a code quantity subject to in-frame compression with a code quantity of inter-frame compression in the unit of macro blocks and a coding mode selection section 37 decides the coding mode based on a history of the coding mode and the comparison result by the comparison arithmetic section 36 to control the switch 33. Thus, the coding efficiency is improved more than the case of deciding a macro block type in the unit of frames.

Description

Detailed Description of the Invention

[Object of the Invention]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video signal high efficiency coding apparatus using orthogonal transform.

[0002]

2. Description of the Related Art In recent years, high-efficiency image coding has been studied. The high-efficiency coding technique is for coding image data at a low bit rate in order to improve the efficiency of digital transmission and recording. In this high efficiency coding, orthogonal transformation such as DCT (discrete cosine transform) processing is performed in small block units with m × n pixels as small blocks. Orthogonal transformation is to transform an input sample value into an orthogonal component such as a spatial frequency component. This makes it possible to reduce spatial correlation components. The components subjected to the orthogonal transformation are quantized to reduce the redundancy of the signal of the small block.

Further, variable length coding such as Huffman coding is applied to the quantized output to further reduce the data amount. Huffman coding performs coding based on the result calculated from the statistical code amount of the quantized output, assigning short bits to data with a high appearance probability and long bits to data with a low appearance probability. The variable length coding to be assigned reduces the total amount of data.

As described above, since variable length coding is adopted in high efficiency coding, if an error occurs even in 1 bit on the way, the code synchronization is lost and it becomes impossible to decode the subsequent data, so that the picture quality is improved. Is significantly deteriorated. Further, in the variable length coding, the compression code amount differs depending on the pattern. Therefore, when recording the compressed data in the storage medium, feedback control is performed according to the pattern in order to make the code amount constant for each frame. However, if there is a difference in the fineness of the picture between the first half and the second half of the screen, the position on the screen of the compressed data cannot be grasped from the recording position.

As a solution to these problems, in Japanese Patent Laid-Open No. 4-91587, one large block is formed by sample data at a plurality of discrete positions on the screen, and the code amount is made constant in units of this large block. A device for doing so has been proposed.

FIG. 11 is a block diagram showing this proposal.

The input digital video signal is given to the large block forming section 14. The large block forming unit 14 forms an input video signal into a frame and then into a large block. That is, the image data of each pixel is divided into macroblock units, which are a group of horizontal and vertical small blocks of 8 × 8 pixels, and as shown in FIG. Part) constitutes a large block. In FIG. 12, the screen is divided into five parts I0 to I4,
Collect one macroblock of each part I0 to I4 and
An example of constructing two large blocks is shown. Other large blocks are similarly constructed. Since each small block is shuffled to form each large block, the amount of information in each large block is considered to be substantially the same regardless of the pattern.

The data of a large block is supplied to the small block forming unit 15 and divided into small blocks in units of 8 × 8 pixels, and then supplied to the orthogonal transformation unit 16. The orthogonal transformation unit 16 orthogonally transforms the input data in small block units and outputs the transform coefficient to the quantization unit 20 via the buffer 17 and also to the data length estimation unit 18.

The quantizing unit 20 has a plurality of quantizers for quantizing with a plurality of quantizing widths for each band, and uses a quantizer designated by the quantizer selecting unit 19 for buffering. 17
Quantizes the transform coefficient from and outputs it to the variable length coding unit 21. The data length estimation unit 18 calculates the data amount when the transform coefficient is quantized with a predetermined quantization width for each large block, and outputs it to the quantizer selection unit 54.

This amount of data can be controlled by changing the quantization width. For example, if the quantization width is made coarse, the information deteriorates, but the dynamic range of the quantized output becomes small and the code amount also becomes small. Quantizer selection unit 19
Selects the quantizer of the quantizer 20 based on the output of the data length estimation unit 18 to make the code amount constant in large block units. The buffer 17 delays the orthogonal transform coefficient until the quantizer of the quantizer 20 is selected.

The quantized output from the quantization unit 20 is given to the variable length coding unit 21, the variable length coding unit 21 variable length codes the quantized output, and outputs the coded output via the transmission unit 22. . In this way, the code amount is made constant for each large block, so the correspondence between the encoded data position and the screen position becomes clear, and even if an error occurs, the data after recovery from the error is reproduced on the screen. Can be used for
Special playback of R is also possible.

FIG. 13 is an explanatory view for explaining a recording method of encoded data of each large block.
This is disclosed in Japanese Patent Publication No. 156792. Figure 1
Reference numerals 3 (a) to 3 (e) show first to fifth sync blocks, respectively.

A large block is composed of five macroblocks as shown by the hatched portion in FIG.
It shall consist of small blocks. For example, in FIG.
Of each of the small blocks I0 to I4 of A1, A2,
... An to E1, E2, ... En. Figure 1
The first sync block shown in 3 (a) is a small block A1.
Through An have a predetermined recording area for recording An. In the first sync block, first, small blocks A1,
Data in the low and middle frequency ranges of A2, ... An are preferentially assigned. Then, if there is a margin in the recording area of each small block,
Record up to the high-frequency data. Further, when there is a margin in the recording area of each small block, the high frequency data of another small block is recorded. Similarly, in the second to fifth sync blocks shown in FIGS. 13B to 13E, small blocks B1, B2, ..., Bn to E1, E2 ,.
Record n data. After the recording of the small blocks corresponding to the respective sync blocks, if there is a free area, the high frequency data of other small blocks that have not been recorded is recorded.

As described above, for the data in the low and middle frequencies,
Since the position of the small block on the screen corresponds to the recording position of the sync block, even if an error occurs, V
Even if all the recorded data of a predetermined sync block is not reproduced as in the case of high-speed reproduction of TR, it is possible to reproduce the screen using the effective reproduction data after recovering from the error.

However, the apparatus of FIG. 11 is a method of shuffling small blocks within one image, and it is not possible to adopt interframe compression as it is. For this reason, when high compression is required, such as when compressing an image with a large amount of information on the screen such as high-definition television broadcasting, the quantization width is extremely large in order to sufficiently reduce the data amount. However, there is a problem that the quality of the reproduced image is deteriorated.

In an apparatus for performing high-efficiency coding between frames, as described in Reference 1 (Hiro Yasuda, "International Standard for Multimedia Coding" Maruzen), MPEG is used.
The hybrid method, which is being studied by (Moving Picture experts group) and the like, has become the mainstream. In this method, in addition to intraframe compression for DCT processing an image in a frame, correlation between frames is used to intercode only a prediction error between an image to be coded and a predetermined reference image. Compression (predictive coding) is also adopted. Note that normally,
Motion-compensated inter-frame prediction that reduces prediction error by motion-compensating a reference image is adopted.

That is, according to this method, an intra-frame coded frame (hereinafter referred to as an I picture) in which one frame of image data is coded by the DCT process and an inter-frame coded frame (hereinafter referred to as a P picture) or I A P-picture in which image data is coded by predictive coding using a picture and an intra-frame, forward, backward, and bidirectional prediction adaptive switching frame (hereinafter referred to as B-picture) constitute a coding frame.

FIG. 14 is an explanatory diagram for explaining the compression method of this system. Since the I picture is coded only by the information in the frame, it can be decoded only by the single coded data. Therefore, in order to prevent error propagation and the like, as shown in FIG. 14, one I picture is inserted in a fixed cycle (for example, 12 frames).

As described on page 139 of Document 1, basically, each prediction method is determined for each frame. Further, as described above, the coding unit is a macroblock (MB), and each MB is classified into four types of intra MB, forward prediction MB, backward prediction MB, and bidirectional prediction MB according to the coding method. Intra MB is M for which only intra-frame coding is performed without inter-frame difference prediction.
B. The forward prediction MB is a MB that has been subjected to motion compensation predictive coding with a reference image in the time axis direction, and the backward prediction MB has been subjected to motion compensation predictive coding with a reference image in the time axis reverse direction. And the bidirectionally predicted MB is
Interpolation of bidirectional reference images on the time axis
n) is the MB for which motion compensation predictive coding is performed.

By the way, in MPEG, an MB type that can exist is determined for each picture type. That is, as shown in Table 1 below, only intra MBs are used for I pictures, and intra MBs or forward prediction MBs are used for P pictures. For B pictures, intra MB, bidirectional prediction MB, forward prediction MB, and backward prediction M.
B is used. Among the MB types, a type having a high coding efficiency, that is, an MB type determined to have a small number of coding bits by calculation or prediction is selected. For example, I
In the 2-frame intra compression in which pictures and P pictures are alternately arranged, all macroblocks of an I picture are intra MBs, and each macroblock of a P picture adaptively selects an intra MB and a forward prediction MB.

[0021]

[Table 1] Therefore, in this case, the MB type is selected only in the P picture, and in the I picture, it is always the intra MB.
Fixed to. Therefore, sufficient encoding efficiency cannot be obtained, and high image quality cannot be achieved by, for example, 2-frame intra compression.

[0022]

As described above, conventionally, a device in which the code amount is made uniform in large block units by shuffling does not support predictive coding (interframe compression), and the image is extremely deteriorated. There was a problem that it would end up. In addition, when the macroblock type determination method adopted in MPEG or the like is adopted in inter-frame compression,
There is a problem that encoding efficiency is low and sufficient image quality cannot be obtained.

It is an object of the present invention to provide a video signal high-efficiency coding apparatus capable of enhancing the coding efficiency by determining the coding mode in units of macroblocks and improving the quality of reproduced images. To do.

[Constitution of Invention]

A video signal high efficiency coding apparatus according to the present invention codes input image data in a coding mode of intra-screen compression or predictive compression for each macroblock which is a coding unit in a frame. Compression means for converting, a storage means for storing the history of the encoding mode of the compression means, and a history of the encoding mode is given to select the encoding mode of the intra-frame compression at least every predetermined number of frames, When a predetermined macroblock of the previous frame is coded in the coding mode of intra-frame compression, a selection means for selecting a coding mode having high coding efficiency among intra-frame compression and predictive compression is provided. When applied to the large block fixed length method, the in-screen compression area and the prediction compression area are set in the screen, and the positions of these in-screen compression area and prediction compression area are surrounded. Area control means that corrects the intra-screen compression area and the prediction compression area for each macroblock that is a coding unit within the screen, and a large block that is configured by a plurality of sample data within the screen. Blocking means, and if the large block is divided into a plurality of conversion units and the conversion unit is based on the intra-screen compression area, orthogonal transformation is performed as it is, and if it is based on the prediction compression area, the prediction error is orthogonalized. The orthogonal transformation means for transforming and outputting the transformation coefficient, and when the predetermined macroblock of the previous frame is included in the intra-frame compression area, the correction of the area control means is controlled to improve the coding efficiency of the orthogonal transformation means. Selecting means for selecting a high conversion mode; and quantizing a conversion coefficient based on the intra-screen compression area with a quantization width according to the intra-screen compression, based on the prediction compression area. Quantizing means for quantizing the transform coefficient with a quantization width according to predictive compression, and controlling the quantization width to make the code amount of the encoded output based on the output of the quantizing means constant in units of the large blocks. When recording, an in-screen compression area and a prediction compression area are set in the screen, and the positions of the in-screen compression area and the prediction compression area are cyclically set. And the data of the intra-screen compression area and the prediction compression area that have been corrected by correcting the intra-screen compression area and the prediction compression area for each macroblock that is a coding unit in the screen based on the coding efficiency. Encoding means for encoding by intra-frame compression or predictive compression, and the encoded output of this encoding means is assigned to a plurality of sync blocks, and the output based on the intra-frame compression is reduced to a predetermined recording range of the sync block. Recording is performed sequentially from the range component, and the high range component of the output based on the intra-screen compression and the low mid range component of the output based on the predictive compression that are left unrecorded are recorded to other unused recording ranges, and then used. The recording means for recording the high frequency component of the output based on the predictive compression is provided in another recording range which is not recorded.

[0025]

In the present invention, the compression means encodes the input image data in macroblock units in the intra-picture compression or predictive compression encoding mode. The storage means stores the history of the coding mode, and the selection means selects the coding mode of the compression means based on the history of the coding mode and the coding efficiency. That is,
The selecting means selects a coding mode having a high coding efficiency from the intra-frame compression and the predictive compression when the predetermined macroblock of the previous frame is coded in the intra-frame compression coding mode, The coding efficiency is increased and the quality of the reproduced image is improved.

[0026]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a video signal high efficiency coding apparatus according to the present invention.

The input image data is given to the small block forming unit 31, and the small block forming unit 31 outputs the input image data in small block units to the terminal a of the switch 33 and also to the inter-frame predicting unit 32. The output of the inter-frame prediction unit 32 is given to the terminal b of the switch 33. The switch 33 is controlled by the encoding mode selection unit 37. The encoding mode selection unit 37 causes the switch 33 to select the terminal a when the intra-frame compression mode is selected, and supplies the block data from the small block formation unit 31 to the orthogonal transformation unit 16 as it is. Further, the coding mode selection unit 37 causes the switch 33 to select the terminal b when the interframe compression mode is selected, and gives the output of the interframe prediction unit 32 to the orthogonal transformation unit 16. The inter-frame prediction unit 32 outputs a prediction error from the image data of the previous frame, as will be described later.

The output of the switch 33 is given to the orthogonal transformation section 16. The orthogonal transformation unit 16 orthogonally transforms the data in units of small blocks and outputs the orthogonal transformation coefficient to the quantization unit 34. Quantizer 34
Transforms the transform coefficient with a predetermined quantization width. The quantized output of the quantization unit 34 is given to the variable length coding unit 21. The variable length coding unit 21 is adapted to, for example, Huffman code the quantized output and output the variable length coded output.

Further, the quantized output is output to the inverse quantizer 35 to create reference image data for inter-frame compression. The inverse quantizer 35 performs an inverse quantization process on the quantized output, restores the data before quantization, and outputs the data to the inverse orthogonal transformer 36. The inverse orthogonal transformer 36 inversely orthogonally transforms the output of the inverse quantizer 35 to restore the data before orthogonal transformation, and the interframe prediction unit 32
Output to. At the time of inter-frame prediction, a prediction error is obtained by subtracting the image of the previous frame from the image of the current frame as a reference image, and the amount of data is reduced by encoding only the prediction error. That is, the inter-frame prediction unit 32 creates the reference block data by motion-compensating the block data of the previous frame from the inverse orthogonal transformer 36, calculates the difference from the block data of the current frame, and calculates the difference value (prediction value). Error) is output to the terminal b of the switch 33.

The coding mode selection unit 37 causes the switch 33
By switching between, the intra-frame compression (intra compression) and the inter-frame compression (inter compression) are switched.
In the present embodiment, this switching is performed in macroblock units. That is, the encoding mode selection unit 37
Determines whether to perform intraframe (intrascreen) compression or interframe compression (prediction compression) for each macroblock in one screen, and attempts to encode the macroblock type selection method. It is determined by the coding mode of the macroblock up to the previous frame at the same position as the macroblock to be processed.

For example, n-frame intra compression will be described in which all frames other than intra frames are forward prediction frames, and an intra frame appears once in n frames. When a predetermined macroblock is compressed in the interframe compression mode continuously for (n-1) frames up to the previous frame, the macroblock to be encoded is forcibly intraframe-compressed. When the macroblock at the same position in the previous frame is intraframe-compressed, the coding mode having the highest coding efficiency is selected from interframe compression and intraframe compression. That is, focusing on a predetermined macroblock in the screen, the intra MB becomes at least once in n frames, and the intra MB and the forward prediction MB (hereinafter referred to as inter M in other frames).
(Also referred to as B), the compression is performed in a coding mode having a high coding efficiency.

Whether or not the coding efficiency is high, that is, whether or not the code amount is small, the predetermined macroblock is actually intra-compressed and forward-prediction-compressed, and then DCT-processed to obtain the code amount of the transform coefficient. It may be judged from the size of
It is known that the magnitude of the code amount can be substantially discriminated also by the calculation before the DCT processing shown in the following equations (1) and (2).

Var = {(encoding block) − (reference block)} squared (1) Varor = {(encoded block) -1 (macro block average)} squared (2) where , Var <Varor, the inter-frame compression mode is selected.

The calculation of the above equations (1) and (2) is performed by the comparison calculation section.
It is supposed to be done by 36. That is, the comparison calculation unit 36
Is given data necessary for the calculation of the above formula (1) from the inter-frame prediction unit 32, and the small block forming unit 31 gives the above formula (2).
The data required for the calculation of is given and the magnitudes of the calculation results of the equations (1) and (2) are compared. The comparison result is given to the coding mode selection unit 37.

The coding mode selection unit 37 gives the selected coding mode to the MB I / P history memory 38, and
Based on the coding mode up to the previous frame stored in the B I / P history memory 38 and the calculation result of the comparison calculation unit 36, the coding mode is determined according to the above-described rule. The encoding mode selection unit 37 causes the switch 33 to have a terminal a when the I mode is selected as the encoding mode.
When the P mode is selected, the switch 33 is caused to select the terminal b.

Next, referring to Table 2 below, the reason why the coding efficiency is improved by adopting the macroblock type selection method of this embodiment will be described.

Table 2 below shows a 2-frame intra compression (n =
This is for explaining the coding efficiency when the MB type selection method in MPEG and the MB type selection method in the present embodiment are selected when 2) is adopted.
Normally, DC is generated by compressing a predetermined macroblock between frames.
The code amount when the T process (P mode) is selected is smaller than the code amount when the intra compression and DCT process (I mode) is performed. However, when the motion is extremely large or when a scene change occurs, the code amount may be larger in the P mode than in the I mode. In Table 2, the advantageous MB type column indicates which of the code amount when the I mode is coded and the code amount when the P mode is coded is smaller. The conventional method 1 column (a) is for the first frame in the P mode, and the conventional method column 2 (b) is for the first frame in the I mode. Further, columns (c) and (d) of the first and second embodiments show an example in which the first frame is the P mode and an I mode, respectively.

[0038]

[Table 2] In the 2-frame intra system in MPEG, as shown in (a) and (b) of Table 2, a predetermined macroblock becomes an intra MB every other frame. Therefore, (a)
, The 10th frame becomes the intra MB, and in (b), the 1st, 3rd, ..., 9th frame becomes the intra M.
It becomes B. Further, since either the I or P mode can be selected for the P picture, the intra MB is created by selecting the I mode having high coding efficiency in the third frame in (a). Similarly, in (b), the I mode having high coding efficiency is selected in the 10th frame to create the intra MB. That is, in (a) and (b), 11
An advantageous MB type will be selected for 7 of the frames.

On the other hand, in this embodiment, 2-frame intra compression is performed in MB units. That is, the intra MB and the inter MB are switched in units of MB. Then, in the frame next to the frame in which a predetermined macroblock is compressed in P mode, the condition is that the macroblock at the same position must be compressed in I mode, but another frame, that is, a predetermined macroblock, is compressed. In the frame next to the frame compressed in the I mode, the I mode and the P mode can be adaptively selected.

For example, in Table 2 (c), since P mode compression is performed in the first frame, I mode compression must be performed in the second frame. But,
The I-mode compression is possible in the third frame as in the second frame, and the P-mode compression is possible in the fourth frame instead of the I-mode. After all, P for 2 consecutive frames
If the mode is not compressed, either the I mode or the P mode can be selected, and the compression in the mode shown in Table 2 (c) is possible. Also, in (d) of Table 2, P mode compression can be selected in the 11th frame. Thus, the number of times that the advantageous MB type is selected is 7 in the conventional examples (a) and (b), whereas it is 8 in the present examples (c) and (d).

That is, in the present embodiment, for an MB for which I-mode compression is more advantageous, the encoding efficiency is improved by always selecting an advantageous MB type in the next frame. Therefore, the larger the number of MBs in which the I mode is more advantageous than the P mode, the higher the coding efficiency.

The improvement of the coding efficiency in this embodiment will be described using mathematical expressions. Now, in the 2-frame intra compression, the probability that the code amount will be reduced when a predetermined macroblock is compressed in the I mode is PI, and the probability that the code amount will be reduced when compressed in the P mode is PP. The probabilities PI and PP satisfy PI + PP = 1.

In the MB type selection method in MPEG, the intra MB
And inter MB can be selected, the probability of selecting an advantageous macroblock type is 10
It is 0%. Also, for intra frames, the probability of choosing an advantageous MB type is PI. Therefore, in this method, the probability that selection of the MB type is advantageous can be expressed by the following equation (3).

Probability = (1 + PI) / 2 × 100 (%) (3) On the other hand, in the present embodiment, either intra MB or inter MB can select any one of the predetermined two frames. Yes, the probability of choosing an advantageous MB type in this frame is 100%. In addition, the other frame has the inter MB as the optimum MB type.
Only when two frames are consecutive, the disadvantageous I-mode compression is adopted. Therefore, the probability of selecting an advantageous MB type in this frame is given by equation (4) below.

Probability = (1−P P × P P) × 100 (%) (4) Therefore, the probability that the MB type is optimally selected in this embodiment can be expressed by the following formula (5).

Probability = {(1+ (1-PP * PP)} / 2 * 100 = (2-PP * PP) / 2 * 100 (5) The difference between the probability in this embodiment and the probability according to MPEG is calculated. Then, the following equation (6) is obtained: That is, since PP is 1 or less, it is understood from the equation (6) that the probability in this embodiment is higher than the probability by MPEG.

(3)-(5) = {(2-PP * PP) / 2- (1 + PI) / 2} * 100 = 50PP (1-PP) ≧ 0 (6) Note that PP = 0,1 , The probability of this embodiment is equal to the probability of MPEG. That is, if PP = 1, that is, if the code amount is smaller in P mode than in I mode in all MBs of all frames, the coding efficiency cannot be improved. However, this condition is extremely rare in natural images, and it is considered that the coding efficiency can usually be improved. On the contrary, it is impossible for a natural image that P P = 0, that is, the amount of code is smaller in the I mode than in the P mode in all frames and MBs. Therefore, the coding efficiency of this embodiment is better than that of the conventional example. Although the two-frame intra compression has been described as an example, other interframe compression methods can be similarly applied.

FIG. 2 is a block diagram showing another embodiment of the present invention. 2, the same components as those in FIG. 1 are designated by the same reference numerals and the description thereof will be omitted. This embodiment is applied to the large block fixed length method proposed in Japanese Patent Application No. 5-74655 filed by the applicant of the present application.
In this embodiment, intraframe compression and forward prediction interframe compression are adopted.

The input digital video signal is given to the large block forming section 14. The large block formation unit 14 forms a large block by a plurality of macro blocks at discrete positions on the screen, and outputs image data to the small block formation unit 15 in units of large blocks. The macro block is composed of, for example, four luminance small blocks and two color difference small blocks.

The small block forming section 15 outputs the data of the large block to the terminal a of the switch 52 in small block units and also to the inter-frame predicting section 53. The output of the inter-frame prediction unit 53 is given to the terminal b of the switch 52. switch
52 is controlled by the encoding mode selection unit 59. The encoding mode selection unit 59 divides the screen into a plurality of parts, determines whether to select the intra-frame compression (I) mode or the inter-frame compression (P) mode for each part, and performs a comparison operation. The coding mode is selected based on the calculation result of the unit 36 and the data from the MB I / P history memory 38. The encoding mode selection unit 59 causes the switch 52 to select the terminal a when the I mode is selected, and supplies the block data from the small block formation unit 15 to the orthogonal transformation unit 16 as it is. Further, the encoding mode selection unit 59 causes the switch 52 to have a terminal b when the P mode is selected.
To output the output of the inter-frame prediction unit 53 to the orthogonal transformation unit.
Give to 16. The coding mode selection unit 59 is a small block generation unit.
If the data from 15 is the data corresponding to the intra-frame compression area described later, the switch 53 is caused to select the terminal a, and if it is the data corresponding to the inter-frame compression area described below, the terminal b is selected. It has become. Note that the inter-frame prediction unit 53 outputs a prediction error with respect to the image data of the previous frame, as will be described later.

The output of the switch 52 is given to the orthogonal transformation section 16. The orthogonal transformation unit 16 orthogonally transforms the data in small block units to obtain orthogonal transformation coefficients by the data length estimation unit 57 and the buffer 17.
Output to. The data length estimation unit 57 has a predetermined quantization table for the intra-frame compression block and a predetermined quantization table for the inter-frame compression block. In the intra-frame compression mode, the data length estimation unit 57 quantizes the transform coefficient using the quantization table for the intra-frame compression block to calculate the code amount. Also, the data length estimation section
In the inter-frame compression mode, 57 quantizes the transform coefficient using the quantization table for the inter-frame compression block to obtain the code amount. The data length estimation unit 57 gives the obtained code amount to the quantizer selection unit 54 in units of large blocks. The quantizer selection unit 54 is configured to select the optimum quantization unit of the quantization unit 57 based on the output of the data length estimation unit 57 in order to suppress a large block to a constant code amount.

FIG. 3 is an explanatory diagram for explaining the transform coefficient from the orthogonal transform unit 16, and Tables 3 and 4 below show the quantization table for intra-frame compression and the quantization frame for inter-frame compression of the quantization unit 58, respectively. It is for explaining the conversion table.

[0053]

[Table 3]

[Table 4] The quantizer 58 includes quantizers 1 to 8, and the quantizers 1 to 8 quantize the transform coefficient using the intra-frame compression quantization table or the inter-frame compression quantization table. Has become.

As described above, the spatial coordinate components from the switch 52 are converted into frequency components by orthogonal transformation. The conversion coefficient has a DC component and an AC component indicating the average of the conversion coefficients, and is arranged from the horizontal and vertical low band to the high band as shown in FIG. The intra-frame compression quantization table is used for the transform coefficient in the intra-frame compression mode, and the quantization width is set for each horizontal and vertical band of the transform coefficient. As shown in FIG. 3, the conversion coefficients are divided into four bands a, b, c, from the horizontal and vertical low band to the high band.
When divided into d, for example, the quantizing means 3 has a quantization width Q1 for the transform coefficients in the bands a and b as shown in Table 1 in the intra-frame compression mode, and a quantum for transform coefficients in the bands c and d. The conversion width is Q2. The quantization widths Q1 to Q4 have a relationship of Q1 <Q4, and the quantization width is set smaller as the conversion coefficient in the lower frequency band. Further, the quantizing means 1 has a smaller quantizing width, and the quantizing means 8 has a larger quantizing width. That is, the code amount of the quantized output increases as the quantizing unit 1 decreases, and the code amount decreases as the quantizing unit 8 decreases.

On the other hand, the quantizing means 1 to 8 use the inter-frame compression quantization table for the transform coefficient in the inter-frame compression mode. The inter-frame compression quantization table has the same structure as the intra-frame compression quantization table, and the quantization widths q1 to q4 are set according to the band of the transform coefficient. The quantization widths q1 to q4 have a relationship of q1 <q4. For example, in the interframe compression mode, the quantizing means 1 quantizes the transform coefficients in all the bands with the quantizing width q1 as shown in Table 2, and the quantizing means 8 performs the quantization in the bands a, b, c. The transform coefficient is quantized with a quantization width q3, and the transform coefficient in the band d is quantized with a quantization width q4.
Quantize with.

Generally, the difference between the quantization widths of the low-frequency component and the high-frequency component of the transform coefficient is set smaller in the inter-frame compression quantization table than in the intra-frame compression quantization table. Often.

In Tables 3 and 4, the combination of the quantization widths Q1 to Q4 is matched with the combination of the quantization widths q1 to q4, but it is not necessary to match them.

The quantized output of the quantizing unit 58 is a variable length coding unit.
Give to 21. The variable length coding unit 21 is adapted to, for example, Huffman-code the quantized output and output the variable-length coded output via the transmission unit 22.

The quantized output is also output to the inverse quantizer 55 in order to create reference image data for inter-frame compression. The inverse quantizer 55 performs an inverse quantization process on the quantized output, restores the data before quantization, and outputs it to the inverse orthogonal transformer 56. The inverse orthogonal transformer 56 inversely orthogonally transforms the output of the inverse quantizer 55 into data before orthogonal transformation, and the interframe prediction unit 53
Output to.

FIG. 4 is a block diagram concretely showing the structure of the inter-frame prediction section 53 in FIG.

At the time of inter-frame prediction, a prediction error is obtained by subtracting the image of the previous frame from the image of the current frame as a reference image, and the amount of data is reduced by encoding only the prediction error. That is, the block data of the previous frame from the inverse orthogonal transformer 56 is given to the adder 60 of the interframe prediction unit 53. At the time of inter-frame prediction, the block data input to the adder 60 is a prediction error,
As will be described later, the adder 60 receives the block data of the previous frame from the switch 61 and adds the two inputs to reproduce the image data of the previous frame and output it to the frame buffer 62.

The frame buffer 62 delays the input block data for one frame period and supplies it to the frame memory 63 as reference block (reference image) data. The frame memory 63 stores the reference block data and supplies it to the predictor 64.

The block data of the current frame is given to the predictor 64 from the small block converting unit 15. The input of the predictor 64 may be a small block, which is a unit for performing orthogonal transformation, or a set unit of a plurality of small blocks. Block data at discrete positions on the screen are sequentially input to the predictor 64. The predictor 64 refers to the reference block data stored in the frame memory 63, determines which position of the reference frame the input block data is most similar to, and creates reference block data. . In this case, the predictor 64 also creates a motion vector indicating which block of the reference frame is the reference block.

The reference block from the predictor 64 is given to the subtractor 65 and also to the adder 60 via the switch 61. The subtractor 65 subtracts the reference block data from the block data of the current frame from the small block forming unit 15 to obtain a prediction error, and outputs it to the terminal b of the switch 52. The subtractor 65 may perform the subtraction in units of a plurality of small blocks. The switch 61 is turned on only during inter-frame prediction, and supplies the reference block data from the predictor 64 to the adder 60. That is, in the intra-frame compression mode, since the block data from the inverse orthogonal transformer 56 does not have a prediction error, the adder 60 outputs it to the frame buffer 62 without adding it to the reference block.

Next, the operation of the embodiment thus constructed will be described with reference to the explanatory view of FIG. FIG. 5 shows a video frame.

The input video signal is converted into a large block in the large block conversion section 14, and further into a small block in the small block conversion section 15. The block data from the small block forming unit 15 is given to the terminal a of the switch 52 as it is, and the inter-frame prediction unit 53 obtains a prediction error from the previous frame and gives it to the terminal b of the switch 52.

The encoding mode selection unit 59 is the first of the input images.
In the frame, the screen is divided into a plurality of parts, and it is determined for each part whether to select the intra-frame compression (I) mode or the inter-frame compression (P) mode. Figure 5,
Regions I0 to I5 and P0 to P5 of 9 and 10 indicate units for switching the compression mode in this case, and I0 to I5 indicate regions for intra-frame compression, and P0 to P5.
Indicates an area for performing inter-frame compression.

Now, in the x-th frame, the large block forming section 14 makes the areas I0 to I2, P3, P4 shown in FIG. 5 (a).
1 macroblock (broken line portion) is collected from 5 macroblocks to form one large block. The encoding mode selection unit 59 causes the switch 52 to select the terminal a when the small blocks in the regions I0 to I2 of the large block of the xth frame are output from the small block formation unit 15. Further, the encoding mode selection unit 59 causes the switch 52 to select the terminal b when the small blocks in the areas P3 and P4 are output from the small block formation unit 15.

The orthogonal transform unit 16 is provided with the small block data of the areas I0 to I2 as it is, and the small block data of the areas P3 and P4 is given only the prediction error from the previous frame. That is, when the small block data of the areas P3 and P4 is input to the inter-frame prediction unit 53, the inter-frame prediction unit 53 compares the reference block of the previous frame with the input small block data and displays the screen of the small block data. By analogy with the upper position, the contents of the frame memory 63 are read and reference block data is created. The subtractor 65 subtracts the reference block data from the small block data to obtain the prediction error, and the switch 52
The signal is output to the orthogonal transformation unit 16 via the terminal b. Orthogonal transformation unit
16 orthogonally transforms the input data, outputs the transform coefficient to the quantizing unit 58 via the buffer 17, and supplies it to the data length estimating unit 57.

The data length estimation unit 57 uses the areas I0 to I2.
Also, the amount of code when the transform coefficients for the regions P3 and P4 are quantized and variable-length coded is obtained, and the data length is output in a large block unit. The quantizer selecting unit 54 causes the quantizing unit 58 to select one of the quantizing units 1 to 8 based on the data length obtained in a large block unit.

When any of the selected quantizing means 1 to 8 receives the transform coefficient corresponding to the region I0 to I2, it quantizes the transform coefficient using the quantization table shown in Table 1, and then, When the transform coefficients corresponding to the areas P3 and P4 are input, the transform coefficients are quantized using the quantization table shown in Table 2 and output to the variable length coding unit 21. The variable length coding unit 21 performs variable length coding on the quantized output and outputs it via the transmission unit 22.

On the other hand, the quantized output is also given to the inverse quantizer 55 to create a reference image for the next frame. Inverse quantizer
Reference numeral 55 inversely quantizes the quantized output, and inverse orthogonal transformer 56 inversely orthogonally transforms the inverse quantized output and provides the result to interframe prediction unit 53.
The inter-frame prediction unit 53 creates reference block data based on the image data of the previous frame and stores it in the frame memory 63 (FIG. 4). As described above, the inter-frame prediction unit 53
The predictor 64 creates a reference block based on the similarity between the reference block stored in the frame memory 63 and the input block, and the subtractor 65 outputs the difference between the input block and the reference block as a prediction error.

As described above, the inter-frame prediction section 53 obtains a prediction error by subtracting the data of the corresponding blocks (blocks at the same position when the motion vector is 0) of the previous frame and the current frame. , The code amount of the prediction error coded output is extremely small.

Next, in the (x + 1) th frame, FIG.
As shown in (a) and (b), regions I0 and I1 in the previous frame (xth frame) adopting I-mode coding.
, I2 become areas P0, P1, P2, and areas P3, P4 adopting P-mode coding become areas I3, I4.
The large block forming unit 14 has areas P0, P1 shown in FIG.
One macroblock (broken line part) is collected from I2 to I4 to form one large block. The coding mode selection unit 37 determines the macroblock type in which the code amount decreases in units of macroblocks based on the comparison result of the comparison calculation unit 36 and the output of the MB I / P history memory 38, and the area For the macroblocks in P0, P1 and P2, select the preferred MB type. Figure 5
(B) shows that there are macroblocks I0 and I1 (enclosed by a broken line) in which the coding in the I mode is advantageous in the regions P0 and P1. It should be noted that the I-mode coding is forcibly performed for the macroblock that has adopted the P-mode coding in the previous frame. When the small block data determined to be encoded in the I mode is supplied to the switch 52, the encoding mode selection unit 37 selects the terminal a, and the small block data determined to be encoded in the P mode is changed to the switch 52. Is applied to select terminal b. Other operations are the same as in the x-th frame.

Further, in the (x + 2) th frame, FIG.
As shown in (b) and (c), the regions P0, P1 and P2 become the regions I0, I1 and I2, and the regions I3 and I4 become the region P.
3 and P4. Furthermore, the coding mode selection unit 37 switches the coding mode in macroblock units. For example, in FIG. 5 (c), regions P0 and P1 (broken line portions) where P mode coding is advantageous in regions I0 and I1 and region I3 (broken line broken lines) where I mode coding is advantageous in region P3. Part)
Is present. The part coded in the P mode in the (x + 1) th frame is forcibly coded in the I mode. Thereafter, in the same manner, the macroblock adopting the P-mode coding is forced to adopt the I-mode coding in the next frame, and the macroblock adopting the I-mode coding is forced to adopt the I, P mode in the next frame. Among them, the compression mode with high coding efficiency is adopted. In this way, the I-mode and the P-mode coexist in each of the five divided regions.

Next, the structure of the sync block by the encoded output will be described with reference to FIGS. 6 to 8. FIG. 6 shows the structure of a sync block by the coded output of the xth frame in FIG. Note that FIG. 6 shows an example in which 5 macroblocks forming a large block are recorded in 5 sync blocks, and it is assumed that each macroblock is composed of m small blocks. Further, FIG. 7 shows a specific data array of the areas I01 and I02 of FIG.

As described above, the large block of the xth frame is composed of the sample macroblocks of the intra-frame compression areas I0 to I2 and the inter-frame compression areas P3 and P4. Ipq in FIG. 6 indicates a recording area in which the encoded outputs of the small blocks q included in the intra-frame compression area Ip are arranged. In these recording areas Ipq, the codes are sequentially arranged from the low-mid range component to the high-range component. On the other hand, Ppq indicates a recording area in which the encoded outputs of the small blocks q included in the inter-frame compression area Pp are arranged. Recording area Ppq
Are sequentially arranged from the low-mid range component to the high-range component of the code, the recording area is half the recording area of Ipq.

When the amount of data in the compression area in one frame is larger than the recording capacity of one sync block, first, the data in the low and middle areas is arranged. That is, in FIG. 6, a recording area of 1 sync is allocated for the low and middle range data of the compression area within one frame, and a 1/2 sync recording is allocated for the low and middle range data of the one frame compression area. Allocating space. Then, the high-frequency data in which the intra-frame compression area is not recorded is arranged in the remaining recording area of the inter-frame compression area. That is, the fourth and the fifth of FIGS.
In the remaining 1/2 sync area of the sync block, the high frequency data of the intra-frame compression area is assigned. Furthermore, the fifth
In the remaining area of the sync block, as shown in FIG. 6 (e), high frequency data of the inter-frame compression areas P3 and P4 are arranged.

For example, as shown in FIG. 7A, in the area I01 of the first sync block, first, the encoded outputs of the small blocks 1 of the intra-frame compression area I0 are sequentially arranged from the low frequency band to the high frequency band. . In the area I02 of the first sync block shown in FIG. 7B, the coded outputs of the small blocks 2 of the intra-frame compression area I0 are arranged in order from the low band. The high-frequency components that cannot be completely written in the intra-frame compression area I02 are shown in FIG.
The second half of the fourth sync block is sequentially recorded.

FIG. 8 shows the structure of a sync block formed by the coded output of the (x + 1) th frame shown in FIG. 5B. FIGS. 8A to 8E show first to fifth sync blocks, respectively.

In the (x + 1) th frame, the inter-frame compression area P is set in the first half recording area of the first to third syncs.
The low and middle frequency components of 0, P1 and P2 are recorded, and the low frequency components of the intra-frame compression areas I3 and I4 are sequentially recorded in the fourth and fifth syncs. In addition, 1 / the latter half of the first to third syncs
The high frequency components of the intra-frame compression areas I3 and I4 are sequentially recorded in the second portion. Then, the unrecorded high frequency components of the inter-frame compression areas P0 to P3 are recorded in the second half of the third sync which is the remaining recording area.

In the (x + 1) th frame, the intra MB for intraframe compression exists in the interframe compression areas P0 and P1. When recording the intra MB portion, the same recording as the first and second syncs in FIG. 6 is performed.

As described above, the low and middle frequency components of the intra MB compressed in the frame are arranged with the highest priority, and the high frequency components of the intra MB which cannot be completely arranged in the designated region and the inter-frame compression are performed. The low and middle frequency components of the inter MB are arranged in the following priority order, and finally the high frequency components of the inter MB are arranged. Intra M, low-mid range component of Intra MB
By arranging the high-frequency component of B, the low-middle-frequency component of inter MB, and the high-frequency component of inter MB in order of priority to configure a sync block, all sync blocks such as high-speed playback of VTR can be obtained. Even when the data of No. 2 cannot be reproduced, it is possible to reproduce a wide area on the screen using the reproduced intra MB data.

As described above, in this embodiment, a large block is formed by a plurality of small blocks at discrete positions on the screen, the data length is obtained in units of large blocks, the quantizing means is selected, and the code amount is set. In addition to homogenizing, the inter-frame prediction unit 53 creates the reference block data by determining the similarity between the input small block and the block data of the previous frame, thereby enabling inter-frame prediction and high compression. ing.

Further, a large block is formed by mixing a plurality of small blocks for intra-frame compression and a plurality of small blocks for inter-frame compression, and the code amount of the large block can be made substantially constant. . Further, the reference block of the inter-frame compression block is made to be the intra-frame compression block by changing the intra-frame compression area and the inter-frame compression area in a two-frame cycle and always making the screen center area the intra-frame compression area. The error can be prevented from propagating to the next frame.

In FIG. 5, the intra-frame compression area and the inter-frame compression area are changed in two frame cycles, but they may be changed in other cycles. Also, FIG.
In the above, the screen is equally divided into five, but other division methods may be used.
For example, as shown in FIG. 9, the central area may be further divided into an inter-frame compressed area and an intra-frame compressed area.

By the way, in the above embodiment, the intra-frame compression area and the inter-frame compression area are set in one frame. However, a predetermined frame of a plurality of frames is constituted by the intra-frame compression area and the remaining frames are It may be composed of an inter-frame compression area. FIG. 10 shows an example in which intraframe compressed frames and interframe compressed frames are alternately provided every other frame.

The x-th frame shown in FIG. 10A is divided into five intra-frame compression areas, and one macro block (hatched portion) in each area constitutes a large block. Further, the (x + 1) th frame shown in FIG. 10B is divided into five inter-frame compressed areas, and one macro block (hatched portion) in each area constitutes a large block. However,
Even in the inter-frame compression area, an inter MB is used for a portion having higher I-mode coding efficiency. Further, as shown in FIG. 10C, the intra MB portion of the previous frame may employ either I-mode or P-mode encoding depending on the encoding efficiency.

In the intra-frame compression and the inter-frame compression,
In general, interframe compression has higher coding efficiency and enables high compression. Therefore, for the x-th frame having a large number of intra MBs, the inter-frame compressed MB is set for the x + 1-th frame
Becomes dominant. Therefore, when the usable code amount is the same, a higher image quality can be obtained in a frame having a large number of inter MBs. Therefore, in order to make the image quality of the frame based on the intra-frame compression and the frame based on the inter-frame compression uniform, it is sometimes preferable to use different code amounts in the frames. In this case, the operation of the quantizer 58 is different from that of the embodiment shown in FIG.

That is, by changing the quantization width of the quantization unit 58 for the same pattern between the frame whose intra-frame compression is the base and the frame whose inter-frame compression is the base,
The interframe compression makes the code amount of the large block of the base frame smaller than the code amount of the large block of the base frame. As a result, the inter-frame compression reduces the code amount of the base frame, and the intra-frame compression allocates the code amount to the base frame, thereby homogenizing and improving the image quality. In this case, the frame buffer 62 of FIG. 4 may be omitted.

The present invention is not limited to the above embodiment. For example, in the above embodiment, the screen is divided into five areas, and sample data is collected from each area to form a large block. The number of screen divisions and the division method are not limited. In addition, in the present embodiment, the reference frame used in the inter-frame prediction is explained as one sheet (2
Frame intra), a plurality of reference frames may be used, and in this case, a plurality of frame memories 63 shown in FIG. 4 may be prepared.

[0092]

As described above, according to the present invention, it is possible to improve the image quality of a reproduced image by increasing the coding efficiency by determining the coding mode in units of macroblocks.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a video signal high efficiency coding apparatus according to the present invention.

FIG. 2 is a block diagram showing another embodiment of the present invention.

FIG. 3 is an explanatory diagram for explaining a conversion coefficient.

FIG. 4 is a block diagram specifically showing a configuration of an inter-frame prediction unit in FIG.

FIG. 5 is an explanatory diagram for explaining the operation of the embodiment.

FIG. 6 is an explanatory diagram for explaining the operation of the embodiment.

FIG. 7 is an explanatory diagram for explaining the operation of the embodiment.

FIG. 8 is an explanatory diagram for explaining the operation of the embodiment.

FIG. 9 is an explanatory diagram for explaining the operation of the embodiment.

FIG. 10 is an explanatory diagram for explaining the operation of the embodiment.

FIG. 11 is a block diagram showing a conventional video signal high efficiency encoding device.

FIG. 12 is an explanatory diagram for explaining a large block.

FIG. 13 is an explanatory diagram for explaining the operation of the conventional example.

FIG. 14 is an explanatory diagram illustrating an MB type determination method in a conventional example.

[Explanation of symbols]

16 ... Orthogonal transformation unit, 32 ... Interframe prediction unit, 33 ... Switch, 36 ... Comparison calculation unit, 37 ... Coding mode selection unit, 38 ... M
B I / P history memory

Claims (3)

[Claims] 1. A compression unit that encodes input image data in a coding mode of intra-frame compression or predictive compression for each macroblock that is a coding unit in a frame, and a history of the coding mode of the compression unit is stored. And a storage unit for selecting the intra-frame compression encoding mode at least every predetermined number of frames given the history of the encoding mode, and encoding a predetermined macroblock of the previous frame in the intra-frame compression encoding mode. A video signal high-efficiency coding apparatus, comprising: a selection unit for selecting a coding mode having a high coding efficiency from the intra-frame compression and the predictive compression. 2. A macroblock which is an encoding unit in a screen while setting an in-screen compression region and a prediction compression region in the screen and periodically changing the positions of these in-screen compression region and the prediction compression region. A region control unit that corrects the intra-screen compression region and the prediction compression region for each, a large block forming unit that forms a large block by a plurality of sample data in the screen, and divides the large block into a plurality of conversion units When the transform unit is based on the in-screen compression area, it is orthogonally transformed as it is, and when it is based on the prediction compression area, it is orthogonal transformation means that orthogonally transforms the prediction error and outputs transform coefficients, and When a predetermined macroblock is included in the in-screen compression area, the correction of the area control means is controlled to cause the orthogonal transformation means to select a transformation mode with high coding efficiency. And a quantization unit that quantizes the transform coefficient based on the intra-screen compression area with a quantization width according to the intra-screen compression and quantizes the transform coefficient based on the prediction compression area with a quantization width according to the prediction compression. And a quantizing control means for controlling the quantizing width to make the code amount of the coding output based on the output of the quantizing means constant in units of the large blocks. Efficiency encoder. 3. An in-screen compression area and a prediction compression area are set in the screen, the positions of these in-screen compression area and prediction compression area are periodically changed, and the in-screen compression area is set based on the coding efficiency. Coding means for correcting the intra-screen compression area and the prediction compression area for each macroblock which is a coding unit, and coding the data of the corrected intra-screen compression area and prediction compression area by intra-screen compression or prediction compression, respectively. The encoded output of the encoding means is assigned to a plurality of sync blocks, and the output based on the intra-frame compression is sequentially recorded in a predetermined recording range of the sync block from the low-frequency component to the uncompressed intra-frame compression. Based on the high-frequency component of the output and the low-to-mid component of the output based on the predictive compression in the other unused recording range, and then based on the predictive compression in the other unused recording range Video signal high-efficiency coding apparatus for characterized by including a recording means for recording the high frequency component of the force.