JPH08317387A - Dynamic image compression encoding system - Google Patents

Abstract

PURPOSE: To reduce image degradation even when inter-frame predictive efficiency is lowered by calculating the rate of a luminance change between frames and performing compression encoding while correcting a target code amount corresponding to that ratio. CONSTITUTION: A luminance signal arithmetic part 9 calculates the sum of luminance signals in a frame and a change rate arithmetic part 10 calculates a rate ΔY of the change in a sum Y of luminance signals. Then, a target code amount correcting part 11 corrects a target code amount Dx generated by a target code amount generating part 3 corresponding to a rate Δx of the change and applies it to a transfer rate control part 6 as Dx'. In this case, the correcting part 11 is provided with a table and when calculating the change rate ΔY, the code amount Dx is corrected and defined as Dx' by calculating a correction coefficient (k) corresponding to the ΔY from the table. Next, at the control part 6, the quantizing parameter of a quantizing part 7 is changed based on the corrected code amount Dx' and compression encoding is performed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture compression coding system for carrying out compression coding of moving pictures.

[0002]

2. Description of the Related Art FIG. 5 is a block diagram of a conventional moving picture compression coding apparatus. Referring to FIG. 5, this moving picture compression coding apparatus is based on an international standard such as MPEG, and has a frame memory 2 capable of storing an input image signal 1 for several frames (several screens). Intra-frame encoding unit 4 that reduces the redundant portion in the frame by encoding by using the correlation between pixels in the frame (in-screen)
An inter-frame predictive coding unit 5 for reducing a redundant part between frames by predictive coding by using a correlation between (screens);
Intra-frame encoder 4 or inter-frame predictive encoder 5
A quantization unit 7 for quantizing the data encoded by the above with a predetermined quantization parameter, a variable length encoding unit 8 for variable length encoding the data quantized by the quantization unit 7, and a target code amount D _The target code amount generation unit 3 for generating _x and the data amount as a result of encoding by the variable length encoding unit 8, that is, the code amount, approaches the target code amount D _x generated by the target code amount generation unit 3. The transfer rate control unit 6 changes the quantization parameter in the quantization unit 7.

Here, the quantizing unit 7 divides the entire DCT transform coefficient obtained as a result of applying DCT (discrete cosine transform) to the coded data, for example, by dividing by a certain value, and by the number of small values. By expressing it, the amount of codes is reduced.

Generally, MPEG defines three types of frames: I frame, P frame, and B frame. Here, the I frame is an intra-frame coded image (Intra coded image), the P frame is an inter-frame forward predictive coded image (Predictive coded image), and the B frame is a bidirectional predictive code. Image (Bidi
It is a rectionally predictive coded image). In addition, GOP (Group o
f Pictures).

FIG. 6A shows an example of arrangement of types of each frame in the GOP. In the example of FIG.
The case where the number N of OP frames is "15" and the period M in which an I frame or a P frame appears is "3" is shown. In this case, a plurality of frames F _{0 to} F ₁₄ in the GOP
A series of compression encoding of a moving image composed of, as shown in FIG. 6 (b), first, performs intra-frame coding for I-frame F _0, then the order of the I frame F ₀ for P frame F ₃ Interframe predictive coding is performed by directional prediction, and then interframe predictive coding is performed by bidirectional prediction from I frame F ₀ and P frame F ₃ for B frames F ₁ and F ₂ , and then P frame F ₆ is performed. Is sequentially performed such that interframe predictive coding is performed by forward prediction from P frame F ₃ .

In this way, I frame, P frame, B
The encoding method differs depending on the frame type, and as a result, the generated code amount also differs depending on the type.
In order to quantize the transfer rate R (bits / second) for each P to be substantially constant (target transfer rate), a target code amount D _x (x = 0) generated from the target code amount generating unit 3 is generated. ，
, 1, 2, 3, ...) Must be changed for each frame F _x .

The ISO / IEC "Te" used for standardization at the International Standardization Conference on Storage Media Video Coding
In st MODEL0 ", the target code amount D _x for a frame F _x, the average value of the quantization parameter of the frame encoded previously generated code amount of the frame previously encoded
(The amount of code coded by the variable length coding unit 8) and GO
It is calculated from the remaining number of frames to be encoded in P and the target transfer rate, and is generated from the target code amount generation unit 3. In FIG. 7, when each frame in the GOP is as shown in FIG. 6A, each frame F _x (x =
An example of the target code amount D _x generated for each 0, 1, 2, ...

Next, the operation of the moving picture compression coding apparatus having such a configuration will be described. First, when the image signal 1 is input to the moving image compression encoding apparatus, the input image signal 1 is temporarily stored in the frame memory 2 capable of storing several frames as image frames F ₀ , F ₁ , F ₂ , ... Each frame F ₀ , F ₁ , F ₂ , stored in the frame memory 2
Are sequentially coded by the intra-frame coding unit 4 or the inter-frame predictive coding unit 5 in the order shown in FIG. When a certain frame F _x is encoded, the transfer rate control unit 6 precedes this by the average value of the quantization parameter of the previously encoded frame and the generation of the previously encoded frame. Code amount (variable length coding unit 8
Code amount), the remaining number of frames to be encoded, and the target transfer rate, the target code amount to be assigned to this frame F _x is calculated, and the target code amount generation unit 3 calculates the target code amount. Generate as a quantity D _x . As a result, the transfer rate control unit 6 causes the generated target code amount D
The quantization parameter in the quantization unit 7 is changed according to _x .

This frame F _x is an intra-frame encoder 4
Alternatively, after being coded by the inter-frame predictive coding unit 5, it is quantized by the quantizing unit 7 by the quantizing parameter set according to the target code amount D _x , and then by the variable length coding unit 8. It is converted and output. In this way, after a certain frame F _x is compression-encoded, the next frame can be encoded in the same procedure, and the compression-encoding is performed so that the code amount of the moving image approaches the target code amount. Can be converted.

By the way, in a moving image such as a movie, the content of the image changes greatly at some point (hereinafter, referred to as a scene change), gradually becomes darker (hereinafter, referred to as a fade-out), or gradually. It becomes brighter (below,
It is often called fade-in). In this case, when there is a scene change, fade-out, or fade-in, the correlation between the frames becomes small, and a larger code amount is generated than in a normal image. In that case, coarse quantization is performed, which causes a problem that the visual image quality is significantly deteriorated. In order to avoid such a problem, for example, in Japanese Patent Laid-Open No. 3-35676, in the case of a scene change, the scene change is detected by calculating the difference between frames at that time, and the encoding for that frame is performed. By switching from interframe predictive coding to intraframe coding, there has been proposed a method of avoiding significant deterioration of image quality.

[0011]

However, in the method disclosed in the above-mentioned publication, the scene change can be detected by calculating the difference between frames and detecting the change in the image content. By comparison, it was not possible to detect the gradual changes in fade-in and fade-out. Furthermore, compared to a scene change in which the image changes in 1/30 seconds, the image quality deterioration at the time of fade-in and fade-out, which changes over about 1 second, is visually conspicuous. Therefore, depending on the method of the above publication, , It was not possible to reduce the visual deterioration of image quality.

According to the present invention, when compressing and coding a moving image,
It is an object of the present invention to provide a moving picture compression coding method capable of reducing image deterioration in a visually noticeable scene with a large change in luminance, that is, a scene in which interframe prediction efficiency decreases.

[0013]

In order to achieve the above object, according to the present invention, basically, when compressing and coding a series of moving images composed of a plurality of frames, the rate of change in luminance between frames is calculated. Is calculated, the target code amount is corrected according to the rate of change in luminance between frames, and compression coding is performed.

In the moving picture compression encoding system according to the present invention,
Since the target code amount is modified and set according to the rate of change in luminance between frames, the target code amount can be adaptively assigned even in a scene in which there is a large change in luminance where image quality deterioration is noticeable. It is possible to reduce deterioration.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an embodiment of a moving picture compression encoding apparatus according to the present invention. In FIG. 1, the same parts as those in FIG. 5 are designated by the same reference numerals.

Referring to FIG. 1, in the moving picture compression encoding apparatus of the present embodiment, the luminance signal calculation unit 9 for calculating the sum Y of the luminance signals of each frame, and the change of the sum Y of the luminance signals between frames. Change rate calculation unit 10 for calculating ratio (change rate) ΔY
And the target code amount D _x generated by the target code amount generating unit 3.
Is corrected in accordance with the change rate ΔY and is given to the transfer rate control unit 6 as D _x '.
Further provided.

Here, the target code amount correction unit 11 is provided with a table as shown in FIG. 2, for example, and the change rate ΔY of the sum Y of the luminance signals between frames is calculated by the change rate calculation unit 1.
When calculated by 0, the correction coefficient k corresponding to ΔY is calculated from the table of FIG. 2, and the target code amount D _x generated from the target code amount generation unit 3 is corrected by the correction coefficient k as shown in the following equation. The code amount is D _x '.

[0018]

[Formula 1] D _x '= kD _x

In the table of FIG. 2, the change rate ΔY is
In other words, it means the time required to change the image,
In many cases, a in FIG. 2 is a scene change that instantaneously changes to a dark image, b is a fade-out, d is a fade-in, and e is a scene change that instantaneously changes to a bright image. Therefore, the target code amount D _x generated from the target code amount generation unit 3 is set to the change rate ΔY of the sum Y of the luminance signals between frames.
By correcting to D _x 'by the correction coefficient k corresponding to, not only scene changes but also scenes with large visually noticeable brightness changes such as fade-in and fade-out, that is, scenes with low interframe prediction efficiency It is possible to reduce image deterioration.

Next, the operation of the moving picture compression coding apparatus having such a configuration will be described with reference to the flowchart of FIG.
In the following, as an example, the input image signal 1 is a digital component signal of a luminance signal and a color difference signal, the frame frequency is 1/30 second, and the encoding method is ISO / IEC11172 “MPEG1”. It is assumed that the intra-frame encoding cycle, that is, the number of GOP frames is "15", and the cycle M in which an I frame or a P frame appears is "3" frame. Also,
It is assumed that the sum Y of the luminance signals and the gradient ΔY of the change are calculated at three-frame intervals from the above-described period M, and not in the bidirectional inter-frame predictive coding frame, that is, the B frame. In other words, for convenience of explanation, the I frame,
It is assumed that the sum Y of the luminance signals and the change gradient ΔY are calculated only for the P frame.

Referring to FIG. 3, first, an appropriate sum Y (i) of luminance signals is initially set in the change rate calculator 10 (step S1). Next, the image signal 1 is input, and the input image signal 1 is input to the frame memory 2 in the order as shown in FIG. 6A, for example, as frames F ₀ , F ₁ , F ₂ , F ₃ , F ₄ ,. Luminance signal calculation is performed only on the luminance signal of the I frame or P frame of the frames F ₀ , F ₁ , F ₂ , F ₃ , F ₄ , ... Give to part 9. That is, it is determined whether or not the frame to be encoded is the I frame or the P frame (step S2), and only when the frame is the I frame or the P frame, the brightness signal is given to the brightness signal computing unit 9.

In the example of FIG. 16A, since the frame to be encoded first is the I frame F ₀ , the luminance signal calculation unit 9 is first _{supplied with} the luminance signal of the I frame F ₀ . In the luminance signal calculation unit 9, the sum of the luminance signals of this I frame F ₀ (the sum of the luminance of each pixel in the I frame F ₀ ) Y (0)
Is calculated (step S3). The sum Y (0) of the luminance signals obtained for the I frame F ₀ is sent to the change rate calculation unit 10 and temporarily stored in the internal memory (not shown) of the change rate calculation unit 10, and the change rate calculation unit 10 In 10, the rate of change between the sum Y of the luminance signals of the previous I frame or P frame and the sum Y of the luminance signals of the current I frame or P frame, that is, the change rate ΔY is calculated (step S4). In this case, the current frame is the first I frame F _0.
Therefore, the change rate calculation unit 10 uses the sum Y (i) of the luminance signals initialized in step S1 as the sum of the luminance signals of the previous I frame or P frame, and the I from the luminance signal calculation unit 9 is calculated. The sum Y (0) of the luminance signals of the frame F ₀ and the sum Y (i) of the luminance signals initially set in the change rate calculation unit 10.
And a change rate (for example, a difference) ΔY (0) with In this way, when the change rate calculation unit 10 obtains the change rate ΔY (0), the change rate ΔY (0) is sent to the target code amount correction unit 11.

On the other hand, in the target code amount generation unit 3, ISO /
Using the method of IEC “Test Model 0”, for example, as shown in FIG. 7, the frames F ₀ , F ₃ , to be encoded,
In accordance with the order of F ₁ , F ₂ , F ₆ , ..., Target code amount D ₀ ,
D ₃ , D ₁ , D ₂ , D ₆ , ... Are sequentially generated and given to the target code amount correction unit 11. In this case, the target code amount D ₀ for the I frame F ₀ is given to the target code amount correction unit 11.

As a result, the target code amount correction unit 11
The correction coefficient k (0) corresponding to the change rate ΔY (0) from the change rate calculation unit 10 is calculated from, for example, the table of FIG. 2, and the calculated correction coefficient k (0) is calculated from the target code amount generation unit 3 in accordance with Equation 1. The target code amount D ₀ of (step S5),
The corrected target code amount D ₀ 'is given to the transfer rate control unit 6. The transfer rate control unit 6 changes the quantization parameter in the quantization unit 7 based on the corrected target code amount D ₀ '(step S6).

The first frame, I in this case
When the correction factor k (0) is determined for the frame F ₀ , this correction factor k (0) is retained for use in compression coding the B frame with the next I or P frame.

In this way, for the first frame, in this case the I frame F ₀ , after the quantization parameter is set based on the corrected target code amount D ₀ ′,
The frame F ₀ is compression-encoded by this quantization parameter (step S7).

After that, the process returns to step S2 again to judge whether the frame to be encoded next is an I frame or a P frame. In this case, since the next frame is the P frame F ₃ , the luminance signal of the P frame P ₃ is sent to the luminance signal calculation unit 9, and the luminance signal calculation unit 9 sums the luminance signals of the P frame F _3. Y (3) is calculated (step S3). The sum Y (3) of the luminance signals obtained for the P frame F ₃ is sent to the change rate calculation unit 10 and temporarily stored in an internal memory (not shown) of the change rate calculation unit 10, and the change rate calculation unit 10 Then, the rate of change between the previous sum Y (0) of the luminance signals of the I frame F ₀ and the present sum Y (3) of the luminance signals of the P frame F ₃ , that is, the change rate (for example, the difference) ΔY.
(3) is calculated (step S4).

On the other hand, in the present case, the target code amount generation unit 3 supplies the target code amount D ₃ to the target code amount correction unit 11. In the target code amount correction unit 11, the correction coefficient k (3) corresponding to the change rate ΔY (3) from the change rate calculation unit 10 is calculated from, for example, the table of FIG.
According to Equation 1 by (3) Fix the target code amount D ₃ from the target code amount generation unit 3 (step S5), and gives the target code amount D ₃ 'after the correction to the transfer rate control section 6. In the transfer rate control section 6, based on the target code amount D ₃ 'after the correction, changing the quantization parameter in the quantization unit 7 (step S6).

The correction coefficient k for the P frame F ₃
When (3) is determined, this correction coefficient k (3) is
It is retained for use in compression coding of B frames between frames or P frames.

In this way, for the P frame F ₃ , the quantization parameter is set based on the corrected target code amount D ₃ ′, and then the frame F ₃ is compressed and encoded by this quantization parameter. (Step S7).

Thereafter, the process returns to step S2 again, and it is determined whether the frame to be coded next is an I frame or a P frame. In this case, since the next frame is the B frame F ₁ , this frame F ₁ is output from the target code amount generation unit 3 using the correction coefficient k (0) obtained in the previous I frame F ₀ . The target code amount D ₁ is corrected (step S8), and the corrected target code amount D ₁ ′ is given to the transfer rate control unit 6. The transfer rate control unit 6 changes the quantization parameter in the quantization unit 7 based on the corrected target code amount D ₁ '(step S6).

Thus, the sum B of the luminance signals is not calculated.
For the frame, the target code amount is corrected using the coefficient k of the I frame or P frame immediately before the encoding order (step S8), and compression encoding is performed (step S7).

Therefore, the frames F ₀ , F ₁ , F ₂ , F ₃ , F
₄ , F ₅ , F ₆ , ... Are frames F ₀ , F ₃ , F ₁ , F ₂ , F ₆ ,
When the compression coding is performed in the order of F ₄ , F ₅ , ..., For the frames F ₀ , F ₁ , and F ₂ , the target code amount D is calculated by the correction coefficient k (0) obtained in the I frame F _{0. 0} ,
D ₁ and D ₂ are D ₀ ′ (= k (0) · D ₀ ), D ₁ ′ (= k (0) · D
_{1), D 2 '(=} k (0) · D 2) used is modified as, also, the frame F _3, F _4, the F _5, the correction calculated by P frame F ₃ coefficient k ( 3), the respective target code amounts D ₃ , D ₄ , D ₅ are D ₃ '(= k (3) · D ₃ ),
It is used after being modified as in D ₄ '(= k (3) · D ₄ ), D ₅ ′ (= k (3) · D ₅ ).

FIG. 4 is a diagram showing a specific example of such processing. Note that, in the example of FIG. 4, as the frame, for convenience of description, only the I frame or the P frame is shown, and the B frame is omitted. In the example of FIG. 4, a scene change occurs in the 3rd and 4th frames, and the 19th to 49th frames occur.
Fade out occurs at the th frame, 63rd to 9th
The case where fade-in occurs in the third frame is shown.

In FIG. 4, when there is no scene change, fade-out, or fade-in, the change rate ΔY of the sum Y of the luminance signals is almost "1", which corresponds to c in the table of FIG. k is calculated as "0.8", so that the target code amount at this time is corrected by the correction coefficient "0.8" and reduced. Further, when a scene change occurs, the change rate ΔY of the sum Y of the luminance signals is very large, and in this case, it corresponds to the condition a or e in the table of FIG. 2 (corresponding to the condition e in the example of FIG. 4). Correction coefficient k is calculated as “1.0”,
Therefore, the target code amount is used as it is without correction. Also, when fade-out and fade-in occur,
The change rate ΔY of the sum Y of the luminance signals is not so large as in the case of the scene change. In this case, b in the table of FIG.
Further, the correction coefficient k is calculated as "1.2" corresponding to the condition of d, and therefore the target code amount is the correction coefficient "1.2".
Modified and increased by.

As described above, in this example, the target code amount is maximized in the case of fade-out and fade-in, the target code amount is not corrected in the case of scene change, and when the brightness change is small, The target code amount is reduced. It should be noted that in the case of c in which the change in luminance is small, the correlation between frames is large and the prediction efficiency is also high. Therefore, in the conventional method, a target code amount more than necessary was often given. However, in this embodiment, by reducing the target code amount when the luminance change is small, fade-in,
It is possible to improve the reproducibility of a scene in which there is a large change in brightness such as fade-out and the deterioration of visual image quality is noticeable.

In the above embodiment, the encoding method is M
Although the description has been given using the method in PEG1, the present invention uses M
The same can be applied to encoding methods other than PEG1.

Further, the conditions in the table of FIG. 2 used in the above-described embodiment, that is, the values of the threshold value and the correction coefficient k, the formula for calculating the corrected target code amount, and the sum of the luminance signals are calculated. The frame period and the like can be changed as appropriate.

[0039]

As can be seen from the above description, according to the present invention, when compressing and coding a series of moving images consisting of a plurality of frames, the ratio of the luminance change between frames is calculated, and the luminance between frames is calculated. Since the target code amount is corrected according to the rate of change and compression encoding is performed, it is possible to reduce image deterioration in a scene in which there is a large visible brightness change, that is, in a scene in which inter-frame prediction efficiency decreases.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of a moving image compression encoding apparatus according to the present invention.

FIG. 2 is a diagram showing an example of a correspondence relationship between a change rate ΔY of a sum of luminance signals between frames and a correction coefficient k.

FIG. 3 is a flowchart for explaining the operation of the moving picture compression encoding apparatus in FIG.

FIG. 4 is a diagram showing a specific processing example of the moving image compression encoding apparatus in FIG. 1.

FIG. 5 is a block diagram of a conventional moving image compression encoding apparatus.

[Fig. 6] Fig. 6 is a diagram for describing an encoding order of each frame in a GOP.

7 is a diagram showing an example of a target code amount for each frame in FIG.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 input image signal 2 frame memory 3 target code amount generation unit 4 intra-frame coding unit 5 inter-frame predictive coding unit 6 transfer rate control unit 7 quantization unit 8 variable length coding unit 9 luminance signal calculation unit 10 change rate calculation Part 11 Target code amount correction unit

Claims (2)

[Claims] 1. In a moving picture compression coding method for compressing and coding a series of moving pictures composed of a plurality of frames, a ratio of luminance change between frames is calculated, and a target code amount is calculated according to the ratio of luminance change between frames. A video compression encoding method characterized by modifying the above and performing compression encoding. 2. In a moving picture compression coding system for compressing and coding a series of moving pictures composed of a plurality of frames, after obtaining the sum of the luminance signals of each frame, the rate of change of the sum of the luminance signals between frames is calculated. Calculate and set the target code amount after frame compression encoding according to the rate of change of the sum of luminance signals between frames, and perform the compression encoding of the next frame based on the target code amount. Video compression coding method characterized by.