JPH0310486A - Moving picture encoder - Google Patents

Abstract

PURPOSE:To improve the coding efficiency and to obtain picture quality preferred visually by applying threshold level processing with a larger threshold level toward higher order discrete cosine transformation coefficient in response to the transmission bit rate of the device. CONSTITUTION:The device consists of a discrete cosine transformation section 1 applying discrete cosine transformation a digitized picture signal or an inter- frame difference of the picture signal, a threshold level processing section 11 applying threshold level processing to a transformation coefficient being the output, a quantizing section 3 quantizing the transformation coefficient subject to threshold level processing, a variable length coding section 4 applying variable length coding to the quantized transformation coefficient, a buffer section 5 storing tentatively the transmitted code and deciding the quantization step width and a threshold level setting section 10 setting the threshold level based on the information. Then the threshold level is set larger toward the higher order discrete cosine transformation coefficient while the transmission bit rate and the buffer residual quantity are controlled. Thus, the coding efficiency is improved and a picture preferred visually is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a color video video encoding device used in video telephones, video conference systems, and the like.

2. Description of the Related Art With the development of video encoding technology, video encoding devices for video telephones or video conference systems have been developed.

For example, Duffle Enchi Che7 (W, H, CHEN
) “Scene Adaptive Coder” (Sc
ene Adaptive Coder, IEEE
Trans.

on Comm,, Vol, COM-32, A 3
A video encoding device using orthogonal transform encoding described in J.D., Mar., 1984) is well known.

Hereinafter, a conventional technique in which two-dimensional cosine transformation is used as orthogonal transformation will be explained using FIG.

In FIG. 5, 1 is a discrete cosine transform section, 2 is a threshold processing section, 3 is a quantization section, 4 is a variable length coding section, and 5 is a transmission buffer section.

The operation of the above configuration will be explained below.

The discrete cosine transform unit 1 converts the image signal 6 inputted in blocks of NXN (N is a positive integer) pixels into the (
1) Perform the discrete cosine transform process shown in equation 1, and transform coefficient 7
is output to the threshold processing section 2.

The threshold processing unit 2 performs threshold processing on the conversion coefficient 7 using the threshold value T instructed by the transmission buffer unit 5, as shown in equation (2), and converts the efficiency conversion coefficient 8 into a quantum output to converting section 3.

1.2. ..., N-1) ...
(1) In order to reduce the types of notifications to be encoded, the quantization unit 3 transfers the efficiency conversion coefficient 8 to the transmission buffer unit 5.
Quantization is performed according to equation (3) using the quantization step width Q specified by , and a quantized transform coefficient 9 is output to the variable length encoder 4.

Here, in general, when the image signal is discrete cosine transformed,
Statistically, low-order coefficients (upper left part of the coefficient matrix) have large values, and high-order coefficients (lower right part of the coefficient matrix) have small values. Therefore, by selecting an appropriate value for the threshold value T in the threshold processing section 2, many of the high-order coefficients of the quantized transform coefficients 9 can be set to zero. To avoid this, the variable-length encoding unit 4 performs variable-length encoding in the order shown in FIG. Encodes the length (run length) of consecutive 00s without converting them into digits. In Figure 6, each black circle is a quantization transform coefficient FTN (u, ■) + (u+
V=0.1, 2. , , N-1), and a series of arrows indicate the encoding order.

The transmission buffer section 5 temporarily holds the code generated by the variable length encoding section 4 and sends it out to the line in accordance with the line speed. The transmission buffer section 5 constantly monitors the remaining buffer amount, and the threshold processing section 2
Alternatively, it is fed back to the quantization unit 3 as a threshold value T or a quantization step width Q, and is controlled so that the buffer amount is always at a constant level.

Furthermore, when configuring a video encoding device that supports multiple transmission bit rates represented by m x basic transmission bit rate (m is a positive integer), the same encoding method is applied and the threshold value T And the quantization step width Q is determined by linear calculation from the value for the basic transmission bit rate.

In the description of this prior art, a method of orthogonally transforming the image signal has been described, but there is also a method of orthogonally transforming the inter-frame difference of the image signal.

Problems to be Solved by the Invention In this way, in conventional technology, by uniformly setting coefficients below a threshold value to 0, both low-order coefficients and high-order coefficients, the amount of information is reduced and encoding efficiency is improved. However, as the threshold increases, especially when the transmission bit rate is low,
Good image quality cannot be expected because information on low spatial frequency components, which have a large effect on human visual characteristics, is also lost. Furthermore, when the transmission bit rate becomes high, even information that could originally be encoded is lost due to threshold processing, resulting in a reduction in image quality.

Means for Solving the Problems The present invention provides an orthogonal transform unit that orthogonally transforms an input image signal divided into minute blocks or a prediction error signal obtained by performing prediction between frames, etc., and transform coefficients into several groups. a threshold setting unit that sets a threshold according to the transmission bit rate of the device for each group; and a threshold processing that replaces conversion coefficients whose absolute value is less than the threshold with 0. By providing a section, the above problem is solved.

According to the above configuration, the present invention increases the run length of the transform coefficients and improves the encoding efficiency by setting a large threshold value for high-order coefficients while controlling the transmission bit rate and the remaining buffer amount. , which provides visually pleasing images at each transmission bit rate.

Embodiment FIG. 1 is a block diagram of a moving image encoding apparatus in an embodiment of the present invention. In FIG. 1, 1 is a discrete cosine transform unit that performs discrete cosine transform on a digitized image signal or an inter-frame difference value of the image signal, and 11 is a discrete cosine transform unit that applies a transform coefficient output from the discrete cosine transform unit 1 to a threshold value. 3 is a quantization unit that quantizes the threshold-processed transform coefficient; 4 is a variable-length encoder that encodes the quantized transform coefficient with variable length code; 5 10 is a transmission buffer section that temporarily stores codes to be transmitted and determines the quantization step width, and 10 is a threshold setting section that sets a threshold value based on information to be described later.

The operation of the above configuration will be described below, assuming that processing is performed using 8×8 pixels as one block. Furthermore, the discrete cosine transform unit 1. Quantization section 3. The operation of the variable length encoder 4° transmission buffer unit 5 is as follows:
Since this is the same as the prior art, the explanation will be omitted.

The threshold setting unit IO determines the position of the conversion coefficient F (u, V) for which threshold processing is performed by the threshold processing unit 11, the quantization step width Q calculated by the transmission buffer unit, and the transmission bit rate. A threshold value T for threshold processing is calculated. The threshold processing unit 11 then compares the threshold T calculated by the threshold setting unit 10 with the conversion coefficient F(u, V), and if the threshold T is larger, the conversion coefficient F( u,
(2) Replace with 0.

The detailed configuration of the threshold processing section 11 will be explained below using FIG. 2. In FIG. 2, 12 is a memory for storing the transform coefficient F(U, V) which is the output of the discrete cosine transform unit 1, and 13 is a memory for storing the transform coefficient F(11, V) temporarily stored in the memo 1J12. 14 is an absolute value circuit that calculates the absolute value of the conversion coefficient F (u, V); 15 is a circuit that calculates the absolute value IF (u, V) of the conversion coefficient and a threshold value T; Comparator to compare, 1
Reference numeral 6 denotes a selector which receives the output of the memo 1J12 and the value 0 as input, and switches the output according to the comparison result of the comparator 15.

The operation of the above configuration will be explained below.

Note that 8×8 conversion coefficients F(u, V) are written in the memory 12. First, the readout order control circuit 13 notifies the memory 12 and the threshold setting unit 10 of the readout order of the discrete cosine-transformed coefficients. Zigzag scanning is used, which is said to be the most efficient when performing run-length encoding of 0.The easiest way to implement zigzag scanning is to sequentially read out a memory in which the scanning order has been written in advance.Third. The figure shows 8×8 size discrete cosine transform coefficients F(u, V), (u, V=0.1.
-..., 7) is an example of the memory contents when reading out. In the memory shown in FIG. 3, the value of U is in the upper 3 bits of the address, and the value of V is in the lower 3 bits in the order of scanning, (u, ■) = (0, 0), (0, 1),
(1, O).

(2,O), (1,1), (0,2), (
0,3)・・・・・・・・・(5,7), (6,7
), (7,6), (7゜7). The output of the memory 12 becomes an input to the absolute value circuit 14 and the selector 16. The absolute value circuit 14 is
Calculate the absolute value IF (u, V)l of the conversion coefficient.

Output to comparator 15. The comparator 15 calculates the threshold value T and the absolute value I of the conversion coefficient, both of which are also notified to the value setting unit 10.
Compare F(u,V)I. The selector 16 selects T> l F (u, V
), it outputs 0, and if T≦IF (u, v)I, it outputs the transformation coefficient F(u, v).

Next, the detailed configuration of the threshold setting section 10 will be explained using FIG. 4. In Fig. 4, 17 is the conversion coefficient F
U, V notified from the threshold processing unit 11 of the group number n for dividing (u, V) into a plurality of groups.
18 is a switch for setting the value of m when the transmission bit rate is expressed as m x basic transmission bit rate; 19 is a switch that reads the value of m as address information and stores the numerical value Cm. 20 is a multiplier for multiplying the output Cm of the memo 1J19 by the quantization step size Q determined by the transmission buffer unit 5;
21 is a multiplier that multiplies the output of the multiplier 20 by the group number n, and 22 is an adder that adds the multiplier 21 and the quantization step size Q.

The operation of the above configuration will be explained below.

The group number determining circuit 17 determines the group number n using, for example, equation (4), but such an operation can be easily realized using a comparator and a selector.

The numerical value Cm is the threshold value T set by the threshold setting unit 10.
A numerical value indicating the ratio of the maximum value T max to the quantization step width Q, and the maximum value of the group number n is n max
When x is calculated using equation (5).

The TmaxO value can be appropriately selected depending on m.

(:m== ((Tmax/Q) -1)/n ma
x-(5) The memo 1J19 stores the numerical value Cm calculated in this manner, and the value of m is read out as address information. The multiplier 20 uses the quantization step width Q and the memory 1 notified from the transmission buffer unit 5.
9 and outputs the result to the multiplier 21. Then, the multiplier 21 multiplies the output of the multiplier 20 by the group number n. Further, the adder 22 adds the output of the multiplier 21 and the quantization step width Q, and notifies the threshold value T to the threshold processing unit 11. In the end, the threshold setting unit 10
) is performing the calculation shown in the formula, u, v, Q,
If a method is used in which the value of the threshold T for the value of m is registered in advance in the memory and read out according to the values of u, v, Q, and m, an even simpler configuration can be obtained.

'p=Qx (1+Cmxmax (u, v))
-(6) [However, max (u, v) is u,
Function that calculates the maximum value of v] Note that if the basic transmission bit rate is 64 kbps, for example, when m = 1, the function is (: m = 1 /
When 28°m=18, it is preferable to set Cm to 20.

Furthermore, although the group number n is calculated using equation (4) in this embodiment, it may be calculated using equation (7). However, in this case, the CmO value also needs to be recalculated according to equation (5).

. =0+, ・・・・・・(7
) As described above, according to this embodiment, the high-order discrete cosine transform coefficients can be thresholded with a large threshold depending on the transmission bit rate of the device, so that the run length of 0 can be reduced. It is possible to increase the encoding efficiency and obtain visually more desirable image quality.

Furthermore, in this embodiment, the threshold value T is u, v.

Although it is determined from the values of m and Q, it is also effective to calculate from only the values of m and Q.

In addition to the effects of the invention, the effects of the present invention include improvements in image quality by performing threshold processing using a threshold that preserves as much of the low frequency components as possible, taking into account human visual characteristics and transmission pit rate. can be measured.

In addition, if the threshold level is raised diagonally to match, for example, zigzag scanning processing, it is possible to improve the resolution of components in diagonal directions, which do not require as much resolution as the horizontal and vertical directions due to human visual characteristics. The threshold value is set to high, and encoding efficiency can be further improved.

[Brief explanation of the drawing]

FIG. 1 is a block wiring diagram of a video encoding device according to an embodiment of the present invention, FIG. 2 is a block wiring diagram of a threshold processing section in the main part of FIG. 1, and FIG. A diagram showing the memory contents for controlling scanning, FIG. 4 is a block wiring diagram of a threshold setting section in the main part of FIG. 1, and FIG. 5 is a block wiring diagram of a conventional video encoding device. FIG. 6 is a conceptual diagram showing an explanation of the scanning order by zigzag scanning. DESCRIPTION OF SYMBOLS 1... Discrete cosine transform unit, 2... Threshold processing unit, 3... Quantization unit, 4... Variable length encoding unit, 5...
...Transmission buffer section, 10...Threshold value setting section, 11
...Threshold processing unit, 12...Memory, 13...
Readout order control circuit, 14... Absolute value circuit, 15.
... Comparator, 16... Selector, 17... Group number determining circuit, 18... Switch, 19... Memory, 20.21... Multiplier, 22... Adder.

Claims (2)

[Claims] (1) An orthogonal transform unit that orthogonally transforms a prediction error signal or an image signal obtained by inter-frame differences of image signals in units of N×M (where N and M are natural numbers); a threshold processing unit that replaces transform coefficients less than a threshold value with 0 for each transform coefficient of the transform coefficient matrix, and a quantizer that quantizes the threshold-processed transform coefficients at intervals of the quantization step width. a threshold setting unit that determines a threshold value for the threshold processing based on the transmission bit rate and the quantization step width; and a variable length encoding unit that encodes the quantized transform coefficients. and a transmission buffer unit that temporarily stores encoded information to be transmitted and determines the quantization step width from the residual amount thereof. (2) The threshold setting unit determines the threshold value for the threshold processing based on the position of the transform coefficient in the transform coefficient matrix, the transmission bit rate, and the quantization step width. The moving image encoding device according to claim 1.