JPH10336649A - Dynamic image encoding method and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform real time variable speed encoding. SOLUTION: Code amount difference data 7 which are the difference of a generation code amount 6 immediately before read from a generation code amount memory 19 and a target code amount 4 immediate before obtained in a multiplier 3 are obtained in a subtractor 5. A conversion table 8 stores a quantization step updating value 9, the quantization step updating value 9 of a large absolute value is outputted in the case that the absolute value of the code amount difference data 7 is large and the quantization step updating value 9 of a small absolute value is outputted in the case that the absolute value of the code amount difference data 7 is small. The outputted quantization step updating value 9 is added with a quantization step 11 read from a quantization step memory 12 in an adder 10 and turned to a present quantization step 13.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding method for determining a quantization step of an orthogonal transform coefficient from a value of a generated code amount.

[0002]

2. Description of the Related Art In a conventional moving picture encoding method, an entire moving picture sequence is encoded with a fixed quantization step.
The generated code amount is checked, and if the total sum of the generated code amounts is less than the target value, the quantization step is set smaller. The method of repeating the conversion was adopted.

[0003]

In the conventional moving picture coding method, it is necessary to obtain an optimum quantization step by repeating coding a plurality of times, so that it can be used for off-line coding processing. However, it could not be used for real-time encoding.

An object of the present invention is to provide a moving picture coding method and apparatus capable of performing variable-speed coding in real time.

[0005]

According to the moving picture coding method of the present invention, the value of the quantization step used for coding the immediately preceding picture is updated according to the difference between the generated code quantity and the target code quantity. When the absolute value of the difference between the generated code amount and the target code amount or the square error is large, the absolute value of the change in the quantization step is set large, and when the difference absolute value between the generated code amount and the target code amount is small. Is to set the absolute value of the change of the quantization step small.

The moving picture coding apparatus according to the present invention further comprises a generated code amount memory for storing the immediately preceding generated code amount, a product of the time information and the average code amount, and a multiplication for obtaining the immediately preceding target code amount. Means, a subtraction means for obtaining code amount difference data which is a difference between the generated code amount stored in the generated code amount memory and the target code amount, and a code amount difference data having an absolute value when the absolute value is large. A conversion table for converting to a large quantization step update value and, when the absolute value of the code amount difference data is small, to a quantization step update value having a small absolute value, and a quantization step for storing the current quantization step A first step of adding a quantization step update value output from the transformation memory and a quantization step stored in the quantization step memory and storing the result in the quantization step memory; Means, and encoder of the current quantization step obtained by the first addition means for encoding processing,
There is provided second adding means for adding the encoded data output from the encoder to the generated code amount stored in the generated code amount memory and storing the added data in the generated code amount memory.

In another moving picture coding apparatus of the present invention, a generated code amount memory for storing a generated code amount immediately before, a product of time information and an average code amount are calculated, and a target code amount obtained immediately before is calculated. 1 multiplication means, subtraction means for obtaining code amount difference data which is the difference between the generated code amount stored in the generated code amount memory and the target code amount, and squaring the code amount difference data to obtain a square error The second multiplication means and the code amount difference data
A conversion table for converting an absolute value into a large quantization step update value when the square error is large, and converting the absolute value into a small quantization step update value when the square error is small;
A quantization step memory for storing the current quantization step, a quantization step update value output from the transformation memory and a quantization step stored in the quantization step memory are added, and the result is stored in the quantization step memory. First
, An encoder for encoding the current quantization step obtained by the first addition means, encoded data output from the encoder, and stored in the generated code amount memory. And a second adder for adding the generated code amount and storing the generated code amount in the generated code amount memory.

According to the embodiment of the present invention, if the absolute value of the difference between the generated code amount and the target code amount or the square error is equal to or smaller than the threshold value, the absolute value of the change in the quantization step is set very small. I do.

According to the embodiment of the present invention, if the difference absolute value or the square error between the generated code amount and the target code amount is equal to or less than the threshold value, the quantization step is not changed.

According to the embodiment of the present invention, a different quantization step change width is used depending on whether the difference between the generated code amount and the target code amount is positive or negative.

The present invention modifies a feedback-type fixed-rate encoding method for determining a quantization step value in proportion to a difference between a generated code amount and a target code amount. On the other hand, instead of determining the value of the quantization step, the update amount from the immediately preceding quantization step is determined. FIG. 3 shows the relationship between the coding rate and the distortion in the fixed rate coding method. The magnitude and distortion of the quantization step correspond to 1: 1. In FIG. 3, D is the quantization distortion, q _o average quantization step, R represents the encoding rate, r _o is the average coding rate. Now, the quantization step when a q _o, the image is the result r _p becomes the code amount of the encoding occurs because complex. Here is a r _p> r _o. At this time,
In a normal encoding method, the quantization step is increased and q
Set to _p . Here, q _p > q _o . As a result, the generated code amount decreases toward _ro . Conversely, a quantization step when a q _o, the code amount image is the result r _m of encoded for simple occurred. Where r _m <r _o
It is. In this case, in the conventional coding method, to reduce the quantization step is set to q _m. Here, q _m <q _o . As a result, the generated code amount increases toward _ro .

FIG. 3 shows the relationship between the quantization step and the difference between the generated code amount and the target code amount. In FIG. 4, q _o is an initial value. Generated information volume

[0013]

(Equation 1) Is the integral of the generated code amount r _i per image frame from time 1 to time T, and r _o is the average code usage per frame (target value).

In general, the fixed rate coding method operates so as to approach a constant coding rate. On the other hand, in variable-rate encoding, a large amount of code is given to a complex image,
Conversely, a small code amount can be given to a simple image. In the conventional variable rate encoding method, a series of moving images is encoded using a fixed number of quantization tests,
A code whose sum of generated code amounts is closest to the target code amount is selected. However, this process is equivalent to repeating coding of a moving image from the beginning, and cannot be performed in real time. Therefore, the present invention has devised a technique for controlling in real time. In the coding control method shown in FIG. 2, the quantization step changes even when the difference between the generated code amount and the target code amount is small. Therefore, if the difference between the generated code amount and the target code amount is within a certain range, the quantization step is set so as to hardly change. When the difference between the generated code amount and the target code amount exceeds a set threshold value, the quantization step is increased to suppress the generated code amount. The strength of the suppression is set so as to increase as the difference between the generated code amount and the target code amount increases.

FIG. 5 shows the relationship between the coding speed and the distortion in the variable speed coding control according to the present invention. In FIG. 5, if the difference between the generated code amount and the target code amount is between _rt1 and _rt2 , the quantization step is hardly changed, and when the difference exceeds _rt1 , the quantization step is increased. For example, the value of r _p is increased from q _o to q _p . Thus the difference between the generated code amount and the target code amount, waiting for one dropped to r _t1. If the difference between the generated code amount and the target code amount becomes equal to or less than r _t1, the quantization step is either not changed, or be changed kept to only slightly. Next, consider a case where the difference between the generated code amount and the target code amount becomes a negative value. Difference in generated code amount and the target code amount to reduce the quantization step if the following r _t2. For example, for r _m, to reduce the q _o to q _m.

The change width Δq of the quantization step is made to correspond to the difference between the generated code amount and the target code amount. The new quantization step is q _new = q _old + q. FIG. 6 shows the correspondence between the difference between the generated code amount and the target code amount used for the calculation of the quantization step and the quantization step. When the difference between the generated code amount and the target code amount is small, △ q
Is set to be small, and Δq increases exponentially as the difference increases. As a result, when the generated code amount is not so far from the target value, the quantization step becomes almost close to the fixed value. Conversely, when the generated code amount deviates greatly from the target value, strong feedback can be applied, and the operation is performed so that the difference between the generated code amount and the target code amount becomes equal to or less than a certain value.

Although FIG. 5 shows almost symmetric characteristics with respect to the sign of the difference between the generated code amount and the target code amount, the difference may be asymmetric. Further, in realizing this, a table for outputting Δq for the value of the difference between the generated code amount and the target code amount may be stored in the memory.

Although the present invention has been described with respect to the orthogonal transform coefficients of a moving image, the present invention can also be applied to a coding method using a subband filter when quantizing the output.
Further, the present invention can be used not only in encoding a moving image as a still image for each frame, but also in inter-frame prediction encoding and motion compensation inter-frame prediction encoding.

In the present invention, if the integrated value of the difference between the generated code amount and the target code amount is within a certain range, the operation is almost fixed at the quantization step. Therefore, even if the input image is complicated, the quantization step is performed. The image quality is improved without increasing the size. Also,
When the generated code amount largely deviates from the target code amount, control can be performed such that the generated code amount approaches the target value.

[0020]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of a moving picture coding apparatus according to an embodiment of the present invention.

The moving picture coding apparatus according to the present embodiment comprises a multiplier 3
It has a subtractor 5, a conversion table 8, an adder 10, a quantization step memory 12, an encoder 14, an adder 17, and a generated code amount memory 18.

The conversion table 8 converts the code amount difference data 7 obtained by the subtractor 5 into a large positive quantization step update value 9 when the value of the quantized difference data 7 is a large positive value.
On the other hand, if the value is a small positive value, it is converted to a small positive quantization step update value 9, and if the value of the code amount difference data 7 is negative and the absolute value is large, the absolute value is large negative. When the value of the code amount difference data 7 is negative and the absolute value is small, the quantization step update value 9 is converted to a negative quantization step update value 9 having a small absolute value. The step update value 9 is stored. The quantization step memory 12 stores the current quantization step 1 obtained by the adder 10.
3 is stored. The generated code amount memory 19 stores the generated code amount 18 immediately before obtained by the adder 17.

Next, the operation of this embodiment will be described.

The product of the time information 1 and the average code amount 2 is a multiplier 3
And the target code amount 4 up to immediately before is obtained. On the other hand, the generated code amount 6 up to immediately before is read from the generated code amount memory 19, the generated code amount 6 -the target code amount 4 is calculated in the subtractor 5, and the code amount difference data 7 is calculated.
Is obtained. The code amount difference data 7 is converted to a corresponding quantization step update value 9 in the conversion table 8. The quantization step update value 9 is added to the immediately preceding quantization step 11 in the adder 10, and the current quantization step 13
Becomes The current quantization step 13 is stored in the quantization step memory 12 and input to the encoder 14. The encoding process is executed in the encoder 14, the encoded data amount 16 is output, and the generated code amount 16 per frame is calculated, and is added to the generated code amount 18 up to the present time in the adder 17. Generated code amount memory 1
9 is stored.

FIG. 2 is a block diagram of an encoding apparatus according to another embodiment of the present invention.

In the present embodiment, a multiplier 20 is provided after the subtractor 5 in FIG. 1 to obtain a square error 20 between the generated code amount 6 and the target code amount 4, and as in the first embodiment, 2 Squared error 2
0 and the conversion table 8 according to the sign of the code amount difference data 7
Converts the code amount difference data 7 into a quantization step update value 9.

[0028]

As described above, according to the present invention,
Since real-time variable rate coding is possible, image quality can be improved at the same coding rate as compared to fixed rate coding.

[Brief description of the drawings]

FIG. 1 is a block diagram of an encoding device according to an embodiment of the present invention.

FIG. 2 is a block diagram of an encoding device according to another embodiment of the present invention.

FIG. 3 is a graph showing a relationship between quantization distortion D and coding rate R in fixed rate coding control.

FIG. 4 is a graph showing a relationship between a quantization step and a difference between a generated code amount and a target code amount.

FIG. 5 is a graph showing a relationship between a quantization distortion D and a coding rate R in variable rate coding control.

FIG. 6 is a graph illustrating a relationship between a quantization step and a difference between a generated code amount and a target code amount.

[Explanation of symbols]

 1 Time information 2 Average code amount 3 Multiplier 4 Target code amount 5 Subtractor 6 Code amount generated immediately before 7 Code amount difference data 8 Conversion table 9 Quantization step update value 10 Adder 11 Quantization step immediately before 12 Quantization Step memory 13 Current quantization step 14 Encoder 15 Encoded data 16 Generated code amount per frame 17 Adder 18 Generated code amount up to now 19 Generated code amount memory 20 Multiplier 21 Square error

Claims (6)

[Claims] In a moving picture coding method for determining a quantization step of an orthogonal transform coefficient from a value of a generated code amount, the method is used for coding the immediately preceding image in accordance with a difference between the generated code amount and a target code amount. When updating the value of the quantization step, if the difference absolute value or the square error between the generated code amount and the target code amount is large, the absolute value of the change in the quantization step is set to be large, and the generated code amount and the target A moving picture coding method characterized in that when the difference absolute value of the code amount is small, the absolute value of the change in the quantization step is set small. 2. The method according to claim 1, wherein when a difference absolute value or a square error between a generated code amount and a target code amount is equal to or smaller than a threshold value, the absolute value of the change in the quantization step is set to be very small. Video encoding method. 3. The moving picture coding method according to claim 1, wherein the quantization step is not changed if the difference absolute value or the square error between the generated code amount and the target code amount is equal to or smaller than a threshold value. 4. The moving picture coding method according to claim 1, wherein a different quantization step change width is used depending on whether the difference between the generated code amount and the target code amount is positive or negative. 5. A moving picture coding apparatus for determining a quantization step of an orthogonal transform coefficient from a value of a generated code amount, comprising: a generated code amount memory for storing a generated code amount immediately before; a product of time information and an average code amount. Multiplying means for calculating a target code amount until immediately before; subtracting means for obtaining code amount difference data which is a difference between the generated code amount stored in the generated code amount memory and the target code amount; If the absolute value is large, the amount difference data is converted to a large quantization step update value, and if the absolute value of the code amount difference data is small, the amount difference data is converted to a small quantization step update value. A quantization table that stores the current quantization step; a quantization step update value that is output from the conversion memory; and a quantization table that is stored in the quantization step memory. A first step of adding a quantization step and storing the result in the quantization step memory
An encoding unit that encodes the current quantization step obtained by the first adding unit; and encoded data output from the encoder and stored in the generated code amount memory. A moving image coding apparatus comprising: a second adding unit that adds the generated code amount and stores the generated code amount in the generated code amount memory. 6. A moving picture coding apparatus for determining a quantization step of an orthogonal transform coefficient from a value of a generated code amount, comprising: a generated code amount memory for storing a generated code amount immediately before; a product of time information and an average code amount. First multiplication means for calculating the target code amount immediately before; and subtraction means for obtaining code amount difference data that is a difference between the generated code amount stored in the generated code amount memory and the target code amount. A second square of the code amount difference data to obtain a square error
Converting the code amount difference data into a quantization step update value having a large absolute value if the square error is large, and a quantization step having a small absolute value if the square error is small. A conversion table for converting to an update value, a quantization step memory for storing a current quantization step, and a quantization step update value output from the conversion memory and a quantization step stored in the quantization step memory. Add and store in the quantization step memory
An encoding unit that encodes the current quantization step obtained by the first adding unit; and encoded data output from the encoder and stored in the generated code amount memory. A moving image coding apparatus comprising: a second adding unit that adds the generated code amount and stores the generated code amount in the generated code amount memory.