JPH03140074A - Moving picture coding device - Google Patents

Abstract

PURPOSE:To improve the picture quality by decreasing a quantization step width than that of a buffer coding an area not including a character or complicated contour when the area including the character or complicated contour is included so as to preserve the high frequency component of the spatial frequency. CONSTITUTION:The coder is provided with a discrete cosine transformation section 1, a threshold level processing section 2, a quantization section 3, a variable length coding section 4, a transmission buffer section 5, an area discrimination section 10 and a quantization step width regulation section 11. Even when the buffer residual quantity is the same but the block including a character or complicated contour is included, the block is minutely quantized with a quantization step width by an offset more than the case of a block not including the character or complicated contour. Thus, even to a picture in mixture with a character and a natural picture, the local picture quality deterioration in the character area is prevented.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application Background Art With the development of video encoding technology, video encoding devices for video telephone or video conference systems have been developed.
For example, double, H, CHEN
) et al [Scene Adaptive Korg J 5
scene Adaptive 0order', I E
BETrans, on C! omm,, Vol.
, C0M-32, No, 3. Mar.

A video encoding device using orthogonal transform encoding described in 1984 is well known.

Hereinafter, a conventional technique in which two-dimensional discrete cosine transform is used as the orthogonal transform will be explained using FIG. In FIG. 4, 1 is a discrete cosine transform section, 2 is a threshold processing section, 3 is a quantization section, 4 is a variable length encoding section, and 5 is a transmission buffer section.

The operation of the above configuration will be explained below.

The discrete cosine transform unit 1 performs the discrete cosine transform process shown in equation (1) on the digital image signal 6 input in blocks of NxN (N is a positive integer) pixels, and thresholds the transform coefficient 7. Output to section 2.

, (.v) -40(")0(") 2 The threshold processing unit 2 performs threshold processing as shown in equation (20) using a constant determined according to the transmission bin rate as the threshold T, The efficiency conversion coefficient 8 is output to the quantization unit 3.

F (u, v) − T (F (Ll, V) > T
) FT(u, y) = 0 (F(u, v
)≦T′)−(2),. . 8 (+1 to 2)” N + (u+ v=o+ L 4 “’r N 1)””
``(') In order to reduce the types of notifications to be encoded, the quantization unit 3 converts the efficiency conversion coefficient 8 according to equation (3) using the quantization step width Q instructed from the transmission buffer unit 6. Quantize and convert the quantized transform coefficients 9 into a variable length encoder 4
Output to.

Fr (u, v) F D N (u, v) = INT (+ 0.5)
・-=(3) Generally, when an image signal is subjected to discrete cosine transformation, statistically low-order coefficients (upper left part of the coefficient matrix) become large values, and high-order coefficients (lower right part of the coefficient matrix) become large values. ) will be a small value. Therefore, if the threshold value T in the threshold value processing section 2 is selected to be an appropriate value, many of the high-order coefficients of the quantized transform coefficients 9 can be set to 0. For this reason, the variable-length encoding unit 4 performs variable-length encoding in the order shown in Figure 0 (usually called zigzag scanning), and for quantized transform coefficients whose value is 00, it encodes the coefficients one by one. Zuni 0
Encode the length of consecutive 0s (run length). In Figure 0, each black circle is a quantized transform coefficient F T N (u + v)
+(u, v=o, 1.2. . . -N-1), and a series of arrows indicates the encoding order.

The transmission buffer section 6 temporarily holds the code generated by the variable length encoding section 4 and sends it out to the line in accordance with the transmission bit rate. The transmission buffer section 6 constantly monitors the remaining buffer amount, and feeds back the quantization step width Q determined by the formula .

Q≦Q max 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 general, areas containing complex outlines and characters contain many high spatial frequency components, and when encoding these areas, it is difficult to encode natural image areas. higher resolution is required. However, in the conventional technology as in the case of ``Double'', the quantization step width Q is determined as a variable that depends only on the remaining buffer amount regardless of the image characteristics, so especially at low bit rates such as 64 kbps, the outline High frequency components of the expressed spatial frequency are likely to be truncated.

As a result, the image quality of areas containing complex outlines and characters is improved. A transmission buffer that temporarily accumulates and determines the quantization step width from the buffer remaining amount, and a quantization step width adjustment that adds an offset to the quantization step width calculated by the transmission buffer based on the determination result of the area determination unit. By having a section, the above problem is solved.

According to the above configuration, the present invention makes the quantization step width smaller when encoding an area containing characters or a complex outline than when encoding an area that does not include characters or a complicated outline. It preserves high-frequency components of spatial frequencies and improves the image quality of the same area.

Embodiment FIG. 1 is a block diagram of an embodiment according to the present invention. In FIG. 1, 1 is a discrete cosine transform unit for performing discrete cosine transform on a digitized image signal;
2 is a threshold processing unit that performs threshold processing on the transform coefficients that are the output of the discrete cosine transform unit 1; 3 is a quantization unit that quantizes the threshold-processed transform coefficients; and 4 is a variable length code for converting the quantized transform coefficients. 6 is a transmission buffer unit that temporarily stores the code to be transmitted and determines the quantization step width; 1o is a variable length encoder unit that temporarily stores the code to be transmitted and determines the quantization step width; A region determination section 11 determines whether or not the region determination section 10 determines whether or not the quantization step width is determined by the transmission buffer.

The operation of the above configuration will be described below, assuming that processing is performed using 8×8 pixels as one block. Discrete cosine transform unit 1, threshold processing unit 2
, the quantization section 3, the variable length encoding section 4, and the transmission buffer section 6 are the same as those in the prior art, so their explanations will be omitted.

The area determining unit 10 determines the difference between the maximum value and the minimum value of the luminance signal among the image signals within one block (hereinafter, the difference between the maximum value and the minimum value of the luminance signal within the block to be encoded is abbreviated as maximum luminance difference). ) and compares the result with a predetermined threshold S. [Generally, in order to clearly recognize the outline of characters, etc., a large contrast is required, so the amount of change in the luminance signal is large in areas containing characters or complex outlines. '', blocks with a maximum luminance difference less than the threshold S are determined to be blocks that do not contain characters or complex contours, and blocks with a maximum luminance difference greater than or equal to the threshold S are determined to be blocks containing characters or complex contours. It is determined that the block is The value of the threshold S is determined empirically, but when quantizing an image signal with 8 bits, a value of about 60 to 1oo is good.

If the area determining unit 1° determines that the block does not include characters or complex outlines, the quantization step width adjustment unit 11 adjusts the quantization step width Q calculated by the transmission buffer.
The offset value Fs is added to the quantization step width Qa, and the result is notified to the quantization unit 3 as the adjusted quantization step width Qa. On the other hand, area determination unit 1
If o is determined to be a block containing a character or a complex outline, the quantization step width Q calculated by the transmission buffer is directly notified to the quantization unit 3 as the adjusted quantization step width Qa. The FsO value is empirically determined depending on the transmission bit rate of the device.

A detailed configuration example of the area determining section 10 will be explained using FIG. 2. In FIG. 2, 12 is a latch that temporarily stores the maximum value of the luminance signal in the block, 13 is a comparator that compares the newly read luminance signal with the content temporarily stored in the latch 12, and 14 is a block. 16 is a comparator that compares the newly read luminance signal with the content temporarily stored in the latch 14. 16 is the content temporarily stored in the latch 12 and the latch. 14 is a differentiator that takes the difference from the temporarily stored content, 17 is a latch that temporarily stores the output of the differentiator 16, and 18 is latch 1.
This is a comparator that compares the output of 7 and a threshold value S.

The operation of the above configuration will be explained below.

Before reading out one block of image signals, latch 1
2 and latch 14 are reset. The comparator 13 compares the data 19 temporarily stored in the latch 12 with the newly read luminance signal 2o, and if the newly read luminance signal 20 is larger, the comparator 13 changes the signal 21 to be written to the latch 12. Output.

The comparator 16 compares the data 22 temporarily stored in the latch 14 with the newly read brightness signal 2o, and if the brightness signal 20 read out is smaller, the latch 14
A write signal 23 is output to.

The difference calculator 16 calculates the difference between the data 19 temporarily stored in the latch 12 and the data 22 temporarily stored in the latch 14. When the reading of the luminance signal for one block is completed in this way, the latch 12 temporarily stores the maximum value of the luminance signal, the latch 14 temporarily stores the minimum value of the luminance signal, and the subtractor 16 temporarily stores the maximum value of the luminance signal. The maximum brightness difference is output. 1
At the read end signal 24 indicating the end of reading the luminance signal of the block, the maximum luminance difference is temporarily stored in the latch 17, and the latch 12 and the latch 14 are reset. The comparator 18 compares the maximum luminance difference temporarily stored in the latch 17 with a preset threshold S, and notifies the quantization step width adjustment unit 11 of the comparison result as the determination result for the block.

Next, a detailed configuration example of the quantization step width adjustment section 11 will be explained using FIG. 3. In Figure 3, 26
is a selector, and 26 is an adder.

The operation of the above configuration will be explained below.

The selector 26 outputs the value 0 or the value Fs using the determination result of the area determination unit 100 as a control line. That is, when the area determination unit 1o determines that the block does not include characters or complex outlines, it outputs the value Fs, and when the area determination unit 1o determines that the block does not include characters or complex outlines, it outputs the value Fs. If so, output the value 0. Adder 2
6 is the quantization step width Q calculated by the transmission buffer section 6
The output of the selector 25 is added as an offset, and the result is notified to the quantization unit 3 as the adjusted quantization step width Qa.

As described above, according to this embodiment, it is possible to determine which blocks include characters or complex outlines and which blocks do not include them, with a simple configuration and in real time.

In this embodiment, a method of orthogonally transforming the image signal has been described, but the inter-frame difference of the image signal may also be orthogonally transformed. However, in that case, the image signal is input to the region determining section 10, and the inter-frame difference of the image signal is input to the discrete cosine transform section 1.

Effects of the Invention As described above, according to the present invention, even if the remaining buffer amount is the same, when encoding a block that includes characters or complex contours, a block that does not include characters or complex contours is encoded. It is possible to perform finer quantization with a quantization step width smaller by Fs than in the conventional case. Therefore, even for images containing a mixture of text and natural images, local deterioration in image quality in text areas can be prevented, and the effect is significant.

[Brief explanation of the drawing]

Fig. 1 is a block wiring diagram of a moving image encoding device according to an embodiment of the present invention, Fig. 2 is a block wiring diagram of an area determining section which is a main part of the device, and Fig. 3 is a block wiring diagram of a main part of the device. FIG. 4 is a block wiring diagram of a certain quantization step width adjustment section, FIG. 4 is a block wiring diagram of a conventional moving picture encoding device, and FIG. 5 is a diagram showing an explanation of the scanning order by zigzag scanning. 1... Discrete cosine transform unit, 2... Threshold processing unit,
3... Quantization section, 4... Variable length encoding section, 6...
Transmission buffer section, 10... area determination section, 11... quantization step width adjustment section, 12° 14° 17... latch, 13. 15゜Figure 18...Comparator, 16...Differentiator, 26...Selector, 26...Adder.

Claims (2)

[Claims] (1) The prediction error signal or image signal obtained from the inter-frame difference of the image signal is N×M (N, M are natural numbers)
An orthogonal transformation unit that performs orthogonal transformation on a pixel block basis, an area determination unit that determines whether the block includes a character or a complex outline, and a separate transform coefficient for each transform coefficient matrix obtained by orthogonal transformation. A quantizer that quantizes at intervals of a predetermined quantization step width, a variable length encoder that encodes the quantized transform coefficients, and temporarily stores the encoded information to be transmitted and stores its residual information. a transmission buffer unit that determines a temporary quantization step width from the amount; and a quantization step width adjustment unit that determines the quantization step width by adding an offset to the temporary quantization step width based on the result of the area determination unit. A moving image encoding device comprising: (2) Among the image signals of a block of N×M (N and M are natural numbers) pixels, by comparing the difference between the maximum value and minimum value of the luminance signal with a predetermined threshold, 2. The moving image encoding device according to claim 1, further comprising an area determining unit that determines whether a character or a complicated outline is included.