JP2735001B2 - Image coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding apparatus, and more particularly to a code amount control circuit in an image coding apparatus for performing orthogonal transformation on an input image signal and performing variable length coding at a target code amount or less.

[0002]

2. Description of the Related Art Generally, in an image coding apparatus using orthogonal transform and variable length coding, an orthogonal transform is applied to an input image signal so that a transform coefficient value is biased, and a coefficient value having a high occurrence probability is shortened. By assigning a long codeword to a low coefficient value as a codeword, the code amount is compressed as a whole, and encoding is performed with high efficiency.

In such an image coding apparatus, as a means for controlling the code amount to be equal to or less than the target code amount, control is generally performed by changing a quantization step value. This utilizes the characteristic that when the quantization step value is increased, the appearance probability of the invalid coefficient value is increased, so that the code amount can be reduced, and conversely, when the quantization step value is decreased, the code amount increases.

However, in a variable-length code, it is difficult to estimate the generated code amount because the generated code amount does not change linearly with a change in the quantization step value. In addition, since the characteristics are greatly different depending on the type of input image, the variable length code is actually used. There is a problem that the generated code amount cannot be known unless variable length coding is performed, and it is generally difficult to estimate a quantization step value at which coding can be performed with a target code amount or less.

[0005] A conventional image coding apparatus will be described with reference to the drawings. FIG. 4 is a block diagram showing an example of a conventional image coding apparatus, FIG. 5 is a diagram showing a first characteristic of a conventional code amount control model, and FIG. 6 is a diagram showing a second characteristic of a conventional code amount control model. FIG.

In a conventional image coding apparatus, when it is necessary to perform coding within a certain period of time such as video equipment, a quantization circuit is used in parallel by using a plurality of initial quantization step values as shown in FIG. 211 to 218, generated code amount measuring circuit 22
1 to 228, two or more points closest to the target code amount are selected, modeling is performed using an approximate function such as a linear function or a spline function, and an optimal quantization step value used for encoding is calculated. Such a configuration is taken. FIG. 5 shows an example in which the relationship between the initial quantization step value and the generated code amount is modeled. Here, two points near the target code amount are selected, approximated by a linear function, and an optimal quantization step value that is equal to or smaller than the target code amount is obtained.

An example of this method is described in Document 1: Masahiko Enari and Soichi Kashida, "HDTV Codec" for 60 to 140 Mbps, January 1992, Video Information, 51-58.
It is described in detail in the section.

[0008]

As described above, in the conventional image encoding apparatus, when encoding is performed within a predetermined time, encoding is performed with a stable code amount for various input images having different statistical characteristics. In order to enable this, for example,
In this case, as shown in FIG. 4, it is necessary to perform quantization using about eight types of initial quantization step values and measure the amount of generated codes. Since each of the quantization circuits 211 to 218 and the generated code amount measurement circuits 221 to 228 performs complicated calculations and processes, it is difficult to use a plurality of these circuits.
There is a problem that the entire encoding device is large and expensive.

Further, in such a conventional image coding apparatus, when the set value of the initial quantization step value does not match the characteristics of the input image, as shown in FIG. If the variable length coding process is actually performed using the coding step value, there is a problem that the target code amount is exceeded and the entire screen is not coded or the image quality is significantly deteriorated. It is known that it is difficult to control the code amount with high accuracy especially when the input image is data that has already been subjected to the encoding / decoding processing.

According to the present invention, a code amount characteristic function used for obtaining an optimum quantization step value used for encoding is accurately modeled by performing logarithmic approximation, and a code is accurately obtained with a smaller number of processing devices within a certain time. An object of the present invention is to provide an encoding device for a high-quality image signal by controlling the amount.

[0011]

An image coding apparatus according to the present invention divides an input image signal into blocks each having a small area composed of a plurality of pixels, performs an orthogonal transformation on a block basis, and obtains a transformation result as an output result. In an image encoding device that quantizes a coefficient value and outputs a variable-length codeword corresponding to the quantized coefficient value, two types of preset initial quantization step values and a quantization step value used for quantization of a previous frame are provided. And quantizing means for quantizing each of the transform coefficient values using a total of three types of quantization step values, and measuring the cumulative amount of code when the output value of the quantizing means is subjected to variable-length coding. Code amount measuring means for calculating a comparison value between three types of generated code amounts based on the three types of quantization step values and a preset target code amount;
Adaptive selection means for selecting the two kinds of generated code amounts having the smallest absolute value of the comparison value and the quantization step value at that time; and a logarithmic value of the two kinds of generated code amounts selected by the selection means. Logarithmic value calculating means for calculating the logarithmic value of the quantization step value; and the optimal quantization step value from the logarithmic value of the quantization step value output from the logarithmic value calculating means and the logarithmic value of the generated code amount at that time. Code amount control means having a quantization step calculation means for obtaining

An image coding apparatus according to the present invention divides an input image signal into blocks each having a small area made up of a plurality of pixels, performs orthogonal transform on a block basis, and quantizes a transform coefficient value as an output result. In an image encoding device that outputs a variable-length codeword corresponding to a quantization coefficient value, an optimal quantization
A frame delay circuit for delaying the step value by one frame;
A first quantization circuit for quantizing the transform coefficient value with a preset first initial quantization step value, and quantizing the transform coefficient value with a preset second initial quantization step value A third quantization circuit that quantizes the transform coefficient value with a quantization step value of an output of the frame delay circuit; and a first quantization circuit that quantizes the transform coefficient value. The first, second, and third units measure the accumulation of three types of code amounts when the three types of quantization coefficient values of the outputs of the respective encoding circuits are encoded into variable-length data, and output three types of accumulated code amount values. A third generated code amount measuring circuit for obtaining an absolute value of a difference value between the three types of code amount accumulated values and a preset target code amount, and a code amount accumulated value having the smallest absolute value and quantization at that time step value and the quantization scan at that time a small code amount accumulated value of the second absolute value An adaptive selection circuit for outputting a-up value, the logarithm calculation circuit for calculating a logarithm of the four kinds of output signal of the adaptive selection circuit, 4 of the logarithm calculation circuit
Code amount with a quantization step calculation circuit for calculating the optimum quantization step value encoding is performed by using a type of output performs modeling code amount characteristic the following target code amount, the
A control circuit, wherein the frame delay circuit is configured to
The optimal quantization step calculated by the quantization step calculation circuit.
This is a circuit for delaying the loop value by one frame, and includes this code amount control circuit.

[0013]

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

FIG. 1 is a block diagram showing an embodiment of an image coding apparatus according to the present invention, FIG. 2 is a block diagram showing an example of a code amount control circuit in the present embodiment, and FIG. 3 is a code amount control circuit in the present embodiment. FIG. 6 is a diagram illustrating characteristics of a model.

In FIG. 1, the present embodiment uses an input image signal 1
A block dividing circuit 71 that divides the block data 60 into blocks each including a plurality of predetermined pixels;
An orthogonal transform circuit 72 for performing an orthogonal transform on 1 and calculating a transform coefficient value 162, and a code amount control for determining an optimal quantization step value 163 that can be encoded with a target code amount or less based on the transform coefficient value 162. A circuit 73 and a delay circuit 74 for delaying the period conversion coefficient value 162 for which the optimum quantization step value 163 is calculated and outputting a delay conversion coefficient value 164
A quantization circuit 75 that quantizes the delayed transform coefficient value 164 using the optimal quantization step value 163 and outputs a quantized count value 165, and converts the quantized coefficient value 165 into variable-length data using a Huffman code. And a variable length coding circuit 76 for converting and outputting a variable length code word 166.

Next, the operation of this embodiment will be described with reference to FIG.

First, an input image signal 160 is digitized image data, and is composed of a multi-valued black-and-white image, an RGB primary color signal, luminance and two kinds of color difference signals, etc., and has a small area block composed of a plurality of pixels. The data is divided into data 161 by the block dividing circuit 71. In the orthogonal transformation circuit 72,
The orthogonal transform is performed on the block data 161 to output a transform coefficient value 162. DCT (discrete cosine transform) is generally used as the orthogonal transform means, but other transform means may be used.

The conversion coefficient value 162 is calculated by the code amount control circuit 73.
Then, the optimal quantization step value 163 that can be encoded with the target encoding amount or less is determined and output to the quantization circuit 75.

The delay circuit 74 delays the transform coefficient value 162 while obtaining the optimum quantized schip value 163 in the code amount control circuit 73, and outputs a delayed transform coefficient value 164 to the quantization circuit 75.

The quantization circuit 75 quantizes the delay conversion coefficient value 164 using the optimal quantization step value 163 calculated by the code amount control circuit 73, and sends the quantization coefficient value 165 to the variable length coding circuit 76. Output. Variable length coding circuit 7
In step 6, the quantization coefficient value 165 is converted into variable length data using Huffman coding, and a variable length code word 166 corresponding to the quantization coefficient value 165 is output.

Next, a code amount control circuit according to this embodiment will be described with reference to FIG.

In FIG. 2, the code amount control circuit 73 in this embodiment includes a frame delay circuit 17 for delaying the optimal quantization step value 163 by one frame, and a transform coefficient value 16.
2 is a quantization circuit 10 for performing quantization with a preset initial quantization step value 100 stored in a register 20.
And a quantization circuit 11 for quantizing the transform coefficient value 162 with a preset initial quantization step value 101 stored in the register 20, and a quantization step for converting the transform coefficient value 162 to the output of the frame delay circuit 17. A quantization circuit 12 for performing quantization with a value 102, and quantization coefficient values 103, 104, and 105 of outputs of the quantization circuits 10, 11, and 12, respectively.
The generated code amount measuring circuits 13, 14, and 15, which measure the accumulated code amount when encoding is encoded into variable length data and output the accumulated code amount values 106, 107, and 108;
6, 107, 108 and the absolute value of the difference value between the target code amount previously stored in the register 21 and the code amount cumulative value 109 having the smallest absolute value and the quantization step value 11 at that time.
An adaptive selection circuit 16 that outputs a code accumulation value 111 having the second smallest absolute value and a quantization step value 112 at that time, and a pair for calculating logarithmic values of the four types of output signals of the adaptive selection circuit 16. A numerical value calculating circuit 18 and a quantization step calculating circuit for modeling the code amount characteristic using the four types of outputs of the logarithmic value calculating circuit 18 and calculating an optimal quantization step value 163 for performing coding at a target code amount or less. 19.

Next, the code amount control circuit 7 in the present embodiment
The operation of No. 3 will be described with reference to FIGS.

The transform coefficient value 162 is obtained by delaying the two types of preset initial quantization step values 100 and 101 and the quantization step value 163 used for encoding the previous frame by one frame with the frame delay circuit 17. Using the quantization step value 102, the quantization circuits 10, 11, 12 quantize the transform coefficient value 162 using the quantization coefficient values 103, 10
Generated code amount measuring circuits 13, 14, 15 as 4, 105
Output to

The generated code amount measuring circuits 13, 14, and 15 measure the accumulated code amounts when the quantized coefficient values 103, 104, and 105 are encoded into variable-length data, respectively. The adaptive selection circuit 16 as 108
Output to

In the adaptive selection circuit 16, the code amount accumulated value 10
6, 107, 108 and the target code amount 167, the absolute value of the difference value is obtained, the code amount accumulated value 106 having the smallest absolute value, the quantization step value 110 at that time, and the code accumulated value 111 having the second smallest absolute value. And then the quantization step value 11
2 is selected and output to the logarithmic value calculation circuit 18. The logarithmic value calculation circuit 18 is, for example, a ROM (Read Only Mem).
ory) or the like, and the accumulated code amount 109, 1
The logarithmic value of 11 and the logarithmic value of the quantization step values 110 and 112 are calculated and output to the quantization step calculation circuit 19. Any value such as 2, 10 may be used as the base of the logarithmic value.

The quantization step calculation circuit 19 models the code amount characteristics using the logarithmic values 113 and 115 of the code amount accumulated value and the logarithmic values 114 and 116 of the quantization step value. Is calculated and output to the quantization circuit 75.

FIG. 3 illustrates a model of the code amount characteristic function. As shown in FIGS. 5 and 6, the relationship between the quantization step value and the generated code amount differs depending on the assigned variable-length codeword, but has an exponential relationship as a whole. Therefore, the logarithmic value of the quantization step value is set on the x-axis and the logarithmic value of the generated code amount is set on the y-axis, and the slope and intercept of a linear function passing through two points of the output value of the logarithmic calculation circuit 18 are calculated. The minimum quantization step value whose axis does not exceed the logarithmic value of the target code amount 167 is calculated. By setting the minimum quantization step value to the optimum quantization step value 163 used for encoding, code amount control not exceeding the target code amount 167 is performed.

In the code amount control circuit 73 of this embodiment, two initial quantization step values 100 and 10 are used as the quantization step values.
By using the quantization step value 102 used in the encoding process of the previous frame in addition to 1, when the input image is a moving image, the accuracy in the vicinity of the target code amount can be obtained to improve accuracy. There is .

[0030]

As described above, the present invention divides an input image signal into blocks each having a small area composed of a plurality of pixels.
In an image coding apparatus that performs orthogonal transform on a block basis, quantizes a transform coefficient value as an output result, and outputs a variable-length codeword corresponding to the quantized coefficient value, two types of preset initial quantization are provided. Quantizing means for quantizing each of the transform coefficient values using a total of three types of quantization step values of a step value and a quantization step value used for quantization of the previous frame; Generated code amount measuring means for measuring the accumulation of the code amount at the time of long coding, and a comparison value between the three generated code amounts based on the three quantization step values and a preset target code amount. The smallest of 2
Adaptive selection means for selecting a kind of generated code amount and a quantization step value at that time, and a logarithmic value for calculating a logarithmic value of the two kinds of generated code amounts and a logarithmic value of the quantization step value selected by the selection means An optimum quantization step value having a calculation means and a quantization step calculation means for obtaining an optimum quantization step value from a logarithmic value of a quantization step value output from the logarithmic value calculation means and a logarithmic value of a generated code amount at that time. Is provided, the input image signal is analyzed using three types of quantization step values, and the code amount characteristic function is modeled using a logarithmic value to set the target code amount. There is an effect that code amount control not exceeding can be performed within a certain time.

By reducing the number of quantization circuits and generated code amount measurement circuits required to be about eight in the past to three, it is possible to reduce the size and weight of the apparatus and to provide the apparatus at low cost. .

[Brief description of the drawings]

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

FIG. 2 is a block diagram illustrating an example of a code amount control circuit in the image encoding device according to the present embodiment.

FIG. 3 is a diagram illustrating characteristics of a code amount control model in the present embodiment.

FIG. 4 is a block diagram showing a conventional example.

FIG. 5 is a diagram showing a first characteristic of a code amount control model in a conventional example.

FIG. 6 is a diagram showing a second characteristic of the code amount control model in the conventional example.

[Explanation of symbols]

 10, 11, 12, 75 Quantization circuit 13, 14, 15 Code amount calculation circuit 16 Adaptive selection circuit 17 Frame delay circuit 18 Logarithmic value calculation circuit 19 Quantization step calculation circuit 20, 21 Register 71 Block division circuit 72 Orthogonal transformation circuit 73 code amount control circuit 74 delay circuit 76 variable length coding circuit 101, 102 initial quantization step value 102 quantization step value delayed by one frame 103, 104, 105, 164 quantization coefficient value 106, 107, 108 code amount accumulation Value 109,111 Selected code amount cumulative value 110,112 Selected quantization step value 113,115 Log value of code amount cumulative value 114,116 Logarithmic value of quantization step value 160 Input image signal 161 Block data 162 Conversion Coefficient value 163 Optimal quantization step value 164 Delay Conversion coefficient value 165 Quantization coefficient value 166 Variable length codeword 167 Target code amount

──────────────────────────────────────────────────の Continued on front page (51) Int.Cl. ⁶ Identification code FI H04N 1/41 G06F 15/66 330H

Claims (2)

(57) [Claims] An input image signal is divided into blocks each having a small area made up of a plurality of pixels, an orthogonal transformation is performed for each block, and a transform coefficient value as an output result is quantized to correspond to the quantized coefficient value. In an image encoding device that outputs a variable-length code word, a total of three types of quantization step values, that is, two types of preset initial quantization step values and a quantization step value used for quantization of a previous frame, are used. Quantizing means for quantizing each of the transform coefficient values, generated code amount measuring means for measuring the accumulated code amount when the output value of the quantizing means is subjected to variable length coding, and the three types of quantization steps An adaptive method for determining a comparison value between three types of generated code amounts according to values and a preset target code amount and selecting the two generated code amounts having the smallest absolute value of the comparison value and a quantization step value at that time. Selecting means; Means for calculating a logarithmic value of the two types of generated code amounts selected by the means and a logarithmic value of the quantization step value; and a logarithmic value of a quantization step value output from the logarithmic value calculation means. An image coding apparatus, comprising: a code amount control unit having a quantization step calculation unit for obtaining the optimum quantization step value from a logarithmic value of a generated code amount at that time. 2. An input image signal is divided into blocks each having a small area composed of a plurality of pixels, an orthogonal transform is performed on a block basis, and a transform coefficient value as an output result is quantized to correspond to the quantized coefficient value. In an image encoding device that outputs a variable-length code word, a frame delay circuit that delays an optimal quantization step value by one frame, and quantizes the transform coefficient value with a first initial quantization step value set in advance. First
, And a second predetermined conversion coefficient value.
A second quantization circuit that performs quantization using an initial quantization step value, and a third quantization circuit that performs quantization on the transform coefficient value using a quantization step value of an output of the frame delay circuit.
The accumulation of three types of code amounts when the three types of quantization coefficient values output from the first, second, and third quantization circuits are encoded into variable length data is measured, and the three types of code amounts are measured. A first, second, and third generated code amount measuring circuit that outputs a cumulative value; and an absolute value of a difference value between the three types of code amount cumulative values and a preset target code amount. An adaptive selection circuit for outputting a small code amount accumulated value, a quantization step value at that time, a code amount accumulated value having the second smallest absolute value, and a quantization step value at that time, and the adaptive selection circuit A logarithmic value calculation circuit for calculating the logarithmic value of the four types of output signals, and a code amount characteristic is modeled using the four types of outputs of the logarithmic value calculation circuit, and coding is performed at the target code amount or less. a quantization step calculation circuit for calculating the optimum quantization step value, the A code amount control circuit, the full
A frame delay circuit is calculated by the quantization step calculation circuit.
Delaying the optimum quantization step value by one frame
That a circuit, the image encoding device characterized by comprising the code amount control circuit.