JPH0766982A - Code quantity controller - Google Patents

Abstract

PURPOSE:To improve the coding efficiency by using a deficient code quantity by a quantization device when the quantizer satisfying an object code quantity is selected and using an excess code quantity by a quantization device whose quantization width is smaller by one step so as to approach the code quantity to an object code quantity. CONSTITUTION:A calculation section 32 executes calculation of a code quantity by an input quantization number 26 and the result revises deficient/excess data 28, 30, and the result is outputted to a next stage as deficient/excess data 36, 38. The operation is repeated by four stages to obtain final deficient/excess data 54, 56. That is, let a quantization number 52 decided finally be Qn, a caption 54 indicates deficient data quantity in the case of coding by the Qn and a caption 56 indicates a data quantity in excess when coded by a number Qn-1, A quantization parameter operation section 58 uses deficient/excess data to control a quantization parameter 50 to command the use of a quantization device shifted from the Qn in the unit of small blocks for the selected Qn to attain fine-adjustment of the code quantity.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a code amount control device in a coding device, and more particularly to a code amount control device in a coding device using a variable length code.

[0002]

2. Description of the Related Art FIG. 4 shows a schematic block diagram of a conventional high efficiency coding apparatus. In FIG. 4, 210 is a large block conversion unit, 212 is a small block conversion unit, 214 is an orthogonal transformation unit, 216 is a quantizer 218 is a code amount calculation unit, 220
Is a quantizer selection unit and 222 is a variable length coding unit.

First, the image data is converted into a large block 210.
Type in, where it is divided into several large blocks.
This large block is a collection of some small blocks to be described later, and the encoding is controlled so that the code amount is constant in units of large blocks. Therefore, it is often the case that small blocks shuffled in advance on the screen are collected into large blocks so as to average the information amount of the original image.

The small block formation section 212 divides each large block into several small blocks. This small block is
It becomes a unit of orthogonal transformation to be described later, and 8 vertical pixels and 8 horizontal pixels are often used.

The orthogonal transform unit 214 performs an orthogonal transform on each small block and outputs the result (called a transform coefficient). The orthogonal transform is, for example, the discrete cosine transform (DCT).

The transform coefficient is quantized by the quantizing unit 216, but before that, the code amount calculating unit 218 calculates the final code amount in units of large blocks, and according to the result, the quantization is performed. The quantizer selection unit 220 selects a quantizer (or a quantization table) that can achieve a predetermined code amount, that is, does not exceed the target code amount. Details of the code amount calculation unit 218 will be described later.

The transform coefficient quantized by the quantizer 216 is variable-length coded by the variable-length coding unit 222. Specifically, the transform coefficient in each small block is zigzag-scanned from the low frequency side, run-length encoded, and then Huffman encoded. In this specification, a series of these processes will be referred to as variable length coding.

The operation of the code amount calculator 218 will be described. N
When there are a plurality of quantizers, numbers from 1 to N are assigned in order from a quantizer having a small quantization width in advance, and Q _n (n = 1 to N)
Will identify each quantizer. From the viewpoint of the code amount, the smaller n is, the larger the code amount is. As described above, the problem here is to select one Q _n for each large block in order to suppress the target code amount in units of large blocks.

To determine the appropriate quantizer, for example, 2
Use the min search method. That is, specifically, the central quantizer Q _{N / 2} is selected, and quantization and variable length coding are actually executed to calculate the code amount. When the code amount is smaller than the target code amount, the target quantizer is between Q ₁ and Q _{N / 2} , and the intermediate quantizer Q _{N / 4} repeats the same calculation. On the contrary, when the code amount is larger than the target code amount, the quantizer Q _{3N / 4} repeats the same calculation. In this way
When there are N quantizers, the target quantizer can be determined by log ₂ N times. The process of selecting the quantizer when N = 16 is shown in FIG. In FIG. 7, 310 indicates a selection direction when the code amount is smaller than the target code amount, 312 indicates a selection direction when the code amount is larger than the target code amount, and 314.
Indicates a selection route when Q ₆ is selected.

A concrete circuit block diagram of the code amount calculation unit 218 is shown in FIG. Reference numeral 230 is input data, 232 is a quantization parameter determination unit, 234 is a quantization parameter, 23
6, 242, 246, and 250 are quantization numbers Q _n and 238.
Is a target code amount, and 240, 244, 248 and 252 are calculation units.

For the sake of explanation, assuming that the number of small blocks included in one large block is k, the input data 230 has transform coefficients for k small blocks. The quantization parameter determination unit 232 gives a parameter representing the feature of the block to each of the k small blocks, and outputs it as a quantization parameter 234. This quantization parameter 234 is a quantizer Q for one large block.
_{When n} is selected, all k small blocks included in the large block are not quantized by the same quantizer Qn,
Relatively quantizer shifted from slightly Q _{n (e.g., Q n + 2, Q n} + 1, Q n-1, such as Q _n-2) is controlled so as to apply to the respective small blocks.

Calculation units 240, 244, 248, 252
Is a circuit for performing the above-described one-time calculation, and the calculation unit 2
40 includes input data 230 and a quantization parameter 23.
4, the quantization number 236 and the target code amount 238 are input. The calculation units 240, 244, 248, 252 calculate the code amount for one large block (k small blocks),
Quantizer numbers 242, 246, 25 to be calculated in the next stage
It outputs 0,254.

The final quantizer number Q _n for this large block is obtained by repeating the calculations of the calculation units 240, 244, 248 and 252 several times. In FIG. 5, N = 1
Since it is an example of the case of 6 and it is a binary search, the quantizer number 236 of the first stage is n = 8, and four calculations, that is, four calculation units 240, 244, 248, 252, You can determine the quantizer.

FIG. 6 shows a detailed block diagram of the calculation unit 510. 262 is input data, 264 is a quantization parameter, 266 is an input quantization number, 268 is a quantizer, 27
0 is a quantizer selection unit, 272 is a variable length coding unit, 274
Is a code amount integrating unit, 276 is an actual code amount, 278 is a target code amount, 280 is a subtractor, 282 is a code amount difference, 284
Is a code determination unit, 286 is a code determination result, 294 is a quantization number selection unit, and 296 is an output quantization number.

First, the quantizer selection unit 270 selects one quantizer or quantization table according to the input quantization number 264 and the quantization parameter 266, and the quantization unit 268.
Quantizes the input data 262 with the selected quantizer or quantization table. The input data 262 is the transformation coefficient of one large block, that is, k small blocks, and the quantization parameter 266 is changed in small block units, so the quantizer selection unit 270 selects accordingly. Shift the quantizer or quantization table.

The variable length coding unit 272 performs variable length coding on the quantized data, and the code length is calculated by the code amount integrating unit 27.
Output to 4. The code amount integrating unit 274 integrates the input code length for k small blocks, and calculates the input quantization number 2
Subtractor 280 calculates the actual code amount of the large block for 64.
Output to. The subtractor 280 subtracts the target code amount 278 from the output 276 of the code amount integrating unit 274 to obtain the code amount difference 2
82 is output. When the code amount difference 282 is plus, the code amount is too large with the current quantization number, and when it is minus, the code amount is small.

Therefore, the code determination unit 284 determines the polarity (that is, plus or minus) of the code amount difference 282,
The quantization number selection unit 294 converts the input quantization number 266 into an output quantization number 296 according to the determination result 286 of the code determination unit 284. When the input quantization number is n, when the code amount over n + (N / 2 m + 1), when the code amount under selects ^{n- (N / 2 m + 1} ). N is the number of quantizers, and m is the mth stage of the calculation unit. In the example shown in FIG. 5, N = 16 and m = 1 to 4.
Also, in the final stage calculation unit, the output quantization number is below the decimal point, but due to rounding, when the code amount exceeds n
+1 and n when the code amount is under. These are well-known processes in the binary search method.

[0018]

Problems of the conventional example will be described with reference to FIGS. 8 and 9. All of these figures
The relationship between the code amount by the finally selected quantizer Q _n and the code amount by the quantizer Q _{n-1 which} is smaller by _one and the target code amount is shown. Q respectively
The difference between the code amount of _n and Q _n-1 and the target code amount is shown. That is, Q
_When quantized with _n , the code amount is smaller than the target code amount by a, and when quantized with Q _n-1 , the code amount exceeds b. In the case of FIG. 8, the value of a is small and the coding efficiency is good. However, in the case as shown in FIG. 9, a becomes considerably large, the coding efficiency is poor, and it is desirable to select the quantizer Q _n-1 if possible, but slightly b
Therefore, the quantizer Qn is unavoidably selected because the code amount is excessive.

An object of the present invention is to provide a code amount control device which solves such a problem.

[0020]

A code amount control device according to the present invention is a code amount control device for making a code amount constant in a large block unit composed of a plurality of small blocks, and a plurality of code amount control devices for calculating an encoding amount. The code amount calculating means of each stage, the transmitting means for transmitting the insufficient / excessive code amount with respect to the target code amount in each stage to the subsequent stage, and the calculated insufficient amount and excess code amount of the transmitted amount according to the calculation result of the code amount calculating means. Update means for updating either one, and operates according to the insufficient / excessive code amount obtained at the final stage of the code amount calculation means, operates the quantization parameter for each small block encoding, and generates It is characterized in that the code amount is brought close to the target code amount.

[0021]

By the above means, in addition to the excess code amount in the case of using the quantizing means determined similarly to the conventional example, the excess code amount in the case of using the quantizing means having a quantization width smaller by one stage than that. To get. Then, based on the obtained insufficient code amount and excess code amount, the quantization parameter for each small block is controlled so as to approach the target code amount as a whole.

[0022]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the schematic arrangement of a code amount control apparatus which is an embodiment of the present invention. 10 is input data, 12 is a quantization parameter determination unit, 14 is a quantization parameter, 16, 26, 34, 42 and 52 are quantization numbers Q.
_n and 18 are initial insufficient data amount, 20 is initial excess data amount, 22 is target code amount, 24, 32, 40 and 48 are calculation units, 28, 36, 44 and 54 are insufficient data amount, and 30, 3
8, 46 and 56 are excess data amounts, 50 is a quantization parameter before operation, 58 is a quantization parameter operation unit, and 60 is a quantization parameter after operation.

Most of the operations in the calculation units 24, 32, 40 and 48 are exactly the same as those in the conventional example shown in FIG.
The description will be omitted, and the characteristic operation of this embodiment will be described in detail.

Each of the calculation units 24, 32, 40, and 48 has two inputs and two outputs. Taking the calculation unit 32 of the second stage as an example, first, the code by the input quantization number 26 is used. The amount calculation is performed, and either the insufficient data amount 28 or the excess data amount 30 input from the previous stage is updated according to the result, and is output to the next stage as the insufficient data amount 36 or the excess data amount 38, respectively. Here, the insufficient / excessive data amount represents the difference between the code amount based on the current quantization number and the target code amount.

The above operation is repeated four times to obtain the final insufficient / excessive data amounts 54 and 56. To be precise, when the quantization number 52 finally determined is Q _n , the code 5
_Reference numeral 4 indicates the amount of data that is insufficient for the target code amount when encoded by Q _n , and reference numeral 56 indicates the amount of data that exceeds the target code amount when encoded by Q _n-1 . Figure 8
And correspond to a and b in FIG.

The quantization parameter operating unit 58 operates the quantization parameter 50 (which has the same content as the quantization parameter 14 but is delayed) by using this insufficient / excessive data amount, and a new quantization parameter is added. 60 is output. As described in the conventional example, the quantization parameter 60 is a parameter for instructing to use a quantizer that is relatively shifted from Q _{n in} small block units with respect to Q _n selected in large block units. Therefore, it is possible to finely adjust the code amount by operating some of k.

Next, FIG. 2 shows a detailed circuit configuration of the calculation units 24, 32, 40 and 48. In FIG. 2, 62 is input data, 64 is a quantization parameter, 66 is an input quantization number, 68 is a quantizer, 70 is a quantizer selection unit, 72 is a variable length coding unit, and 74 is a code amount integrating unit. , 76 is an actual code amount, 78 is a target code amount, 80 is a subtractor, 82 is a code amount difference, 84 is a code determination unit, 86 is a code determination result, 88 is an insufficient data amount input, and 90 is an excess data amount input. , 92 is a changeover switch, 94 is a quantization number selection unit, 96 is an output quantization number, 98 is an insufficient data amount output, and 100 is an excess data amount output.

In FIG. 2, the same parts as those of FIG. 6 of the conventional example operate in the same manner, so the parts added in FIG. 2 will be described in detail.

The code amount integrating unit 74 outputs the actual code amount for the input quantization number 66 on the output line 76, as described with reference to FIG. The subtractor 80 subtracts the target code amount 78 from the actual code amount 76 and outputs a code amount difference 82.
The code determination result 86 indicates the code of the code amount difference 82, that is, the polarity.

The switch 92 uses the code determination result 86 to
Either the input insufficient data amount 88 or the excess data amount 90 is updated with the code amount difference 82, and the output lines 98, 1
Output to 00. The amount of data that is not updated is output as it is. For example, when the code amount exceeds, as shown in the figure, the insufficient data amount 88 is output as it is, and the excess data amount 90 is replaced with the value of the overwritten data amount 82 and output.

With this configuration, the target amount of insufficient / excessive data can be obtained on the signal lines 54 and 56 via the four-stage calculating section in FIG. This will be described with reference to FIG.
In FIG. 7, reference numeral 314 indicates a process in which the quantizer Q ₆ is selected. By tracing the arrows in the opposite direction, it is understood that the code amount calculation of Q ₆ and Q ₅ is performed in the fourth and third stages, and the target data amount can be obtained.

[0033] Considering the case of another Q _9, missing data of Q ₉ are obtained in the fourth stage, the first stage excess data amount Q ₈ is obtained. In particular, the amount of excess data in the first stage is 2 to
Since the code amount does not exceed in the fourth stage, it is transmitted to the end without being rewritten in the middle. Similarly, for the other Q _n , either one of the missing / excessive data is not calculated even once, so that the correct result cannot be obtained.

FIG. 3 shows a detailed circuit block diagram of the quantization parameter operating unit 58 shown in FIG. 110 is a quantization parameter input, 112 is a selected quantization number Q _n , 114
Is an insufficient data amount, 116 is an excess data amount, 118 is a divider, 122 is a multiplier, 126 is an adder, 130 is a comparator, 132 is a changeover switch, and 134 is a quantization parameter output.

First, the input lack / excess data amount 11
0 and 112 correspond to FIG. 8 and FIG.
If (a ≧ 0, b ≧ 0), the divider 118 uses a / (a
+ B) is calculated, and the ratio of the current insufficient data amount to the code amount that increases when the quantization number is decreased by 1 is 120
Is output. Decreasing the quantization number by one is equivalent to manipulating all k quantization parameters without changing the quantization number and shifting the quantization number of each small block by one. Therefore, by multiplying this ratio 120 by the multiplier 122 by k, it is possible to know how many of the k parameters should be operated.

The adder 126 operates on the quantization parameter. Specifically, it is assumed here that if the parameter is increased by 1, the quantizer is shifted by one smaller. Of the k quantization parameters 110 that have been input, the number of parameters specified by the output 124 of the multiplier 122 is incremented by 1 in the adder 126 and output as a new parameter at the output 128. .

The comparator 130 determines whether or not the quantization number 112 is 1 or 16, and when the quantization number 112 is 1 or 16, there is the above-mentioned problem, so that the switch 132 outputs the parameter 110 before the operation. Although the number of parameters specified by the output 124 of the multiplier 122 is set to the ratio 120 here, it may actually exceed the code amount.

When the quantization parameter value is increased by the adder 126, which small block is to be used is selected, for example, when there is a Y signal and a color difference signal block, Y
A method is conceivable in which the signal blocks are sequentially arranged, or those having a large quantization parameter value are postponed, and those having a small parameter value are sequentially arranged.

Having described one embodiment of the present invention, here:
The basic idea of the present invention will be explained again. Manipulating all k quantisation parameters of each small block to select a quantizer that is one smaller means that Q _n−1 instead of the quantizing number Q _n of the large block without manipulating the quantizing parameters at all. Is equivalent to the use of a quantizer Q _n.
~ Q _n-1 will be obtained.

[0040]

As can be easily understood from the above description, according to the present invention, the quantizer Q _n satisfying the target code amount is
There when selected, by using an excess code amount due to insufficient amount of code and Q _n-1 by Q _n, it is possible to approach further the target code amount, the coding efficiency is improved.

[Brief description of drawings]

FIG. 1 is a schematic block diagram of a code amount control device according to an embodiment of the present invention.

FIG. 2 is a detailed circuit configuration block diagram of calculation units 24, 32, 40, and 48.

FIG. 3 is a detailed circuit block diagram of a quantization parameter operation unit 58.

FIG. 4 is a schematic configuration block diagram of a conventional high efficiency encoding device.

5 is a schematic circuit block diagram of a code amount calculation unit 218. FIG.

FIG. 6 is a detailed block diagram of calculation units 240, 244, 248 and 252.

FIG. 7 is an example of a quantizer selection process.

FIG. 8 is an explanatory diagram of a relationship between a code amount of the quantizers Q _n and Q _n-1 and a target code amount.

FIG. 9 is an explanatory diagram of the relationship between the code amount of the quantizers Q _n and Q _n-1 and the target code amount.

[Explanation of symbols]

10: Input data 12: Quantization parameter determination unit 14: Quantization parameter 16, 26, 34, 42, 52: Quantization number Q _n 18: Initial insufficient data amount 20: Initial excess data amount 22: Target code amount 24, 32, 40, 48: Calculation unit 28, 36, 44, 54: Insufficient data amount 30, 38, 46, 56: Excess data amount 50: Quantization parameter before operation 58: Quantization parameter operation unit 60: After operation Quantization parameter 62: Input data 64: Quantization parameter 66: Input quantization number 68: Quantization unit 70: Quantizer selection unit 72: Variable length coding unit 74: Code amount integration unit 76: Actual code amount 78 : Target code amount 80: Subtractor 82: Code amount difference 84: Code determination unit 86: Code determination result 88: Insufficient data amount input 90: Excessive data amount input 92: Switching switch 94: quantizer number selection unit 96: Output quantization number 98: missing data amount output 100: excess data amount output 110: quantization parameter input 112: selected quantization number Q _n 114: insufficient data amount 116: Excess Data Quantity 118: Divider 122: Multiplier 126: Adder 130: Comparator 132: Changeover switch 134: Quantization parameter output 210: Large blocking unit 212: Small blocking unit 214: Orthogonal transformation unit 216: Quantizer 218 : Code amount calculation unit 220: quantizer selection unit 222: variable length coding unit 230: input data 232: quantization parameter determination unit 234: quantization parameter 236, 242, 246, 250, 254: quantization number Q _n 238: Target code amount 240, 244, 248, 252: Calculation unit 262: Input data 264: Amount Parameterization 266: Input quantization number 268: Quantization unit 270: Quantizer selection unit 272: Variable length coding unit 274: Code amount integration unit 276: Actual code amount 278: Target code amount 280: Subtractor 282: Code amount difference 284: Code determination unit 286: Code determination result 294: Quantization number selection unit 296: Output quantization number 310: Selection direction when code amount is smaller than target code amount 312: Code amount is more than target code amount Selection direction when larger 314: Selection path when Q ₆ is selected

Claims (1)

[Claims] 1. A code amount control device for making a code amount constant in a large block unit made up of a plurality of small blocks, the code amount calculating means having a plurality of stages for calculating an encoding amount, and a target code amount at each stage. According to the calculation result of the transmission amount means and the transmission means for transmitting the insufficiency / excess code amount to
Update means for updating one of the transmitted insufficient code amount and excess code amount, and each small block is encoded according to the insufficient / excess code amount obtained at the final stage of the code amount calculating means. A code amount control device, characterized in that the generated code amount is brought closer to the target code amount by operating a quantization parameter at that time.