JP2664223B2 - Orthogonal transform coefficient quantization circuit - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a quantization circuit for encoding and transmitting an image by orthogonal transform encoding.

[Background Art] Orthogonal transform coding, in which an original signal is orthogonally transformed, its transform coefficient is quantized, and transmitted, is often used because of its high coding efficiency. When applied to an image signal, the image is first divided into N × N (N: an integer) small blocks, and each block is subjected to an orthogonal transform to be converted into N × N frequency domain coefficients. Is quantized.

[Problems to be Solved by the Invention] If the above-mentioned quantization is made coarse for every block in the same manner, in a block including an edge, a decoded image obtained by inversely transforming a quantized coefficient is obtained. The error is diffused around the edge and becomes noise. This noise is usually detected in a flat area when a block contains a fairly strong edge and a flat part.For example, this noise occurs at a part such as the boundary between a person and the background, and the quality of the decoded image is degraded. Had become a major factor.

SUMMARY OF THE INVENTION An object of the present invention is to provide an orthogonal transform coefficient quantization circuit that can solve the above-described problems of the conventional technique and can improve the image quality of a decoded image.

[Means for Solving the Problems] To achieve the above object, the present invention provides a method for converting an image signal into an N signal.
× N (N = integer) dividing means, an orthogonal transform means for performing a discrete orthogonal transform on each block signal obtained by the dividing means to obtain a transform coefficient, and an orthogonal transform means Quantizing means for quantizing the obtained transform coefficient so that the quantization interval can be changed, and for each of at least a plurality of sub-blocks in the block, the edge of the sub-block is Detecting means for detecting whether the area is a flat area or a flat area, and an edge area / flat area mixed block for extracting a block in which one or more edge area sub-blocks are mixed with the flat area sub-block based on the detection result of the detecting means An extraction unit, and based on an extraction result of the extraction unit, a transform coefficient of a block in which an edge region and a flat region are mixed is smaller than that of another block. Wherein said that and means for controlling the quantization interval of the quantizer to reduction.

[Operation] According to the present invention, for each block of an image signal, for each of at least a plurality of sub-blocks in the block, the sub-block is detected as an edge area or a flat area, and the detection result is obtained. A block where one or more sub-blocks of the edge region are mixed with the sub-block of the flat region is extracted based on Become

[Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 shows a first embodiment of the present invention, in which an image signal input from one input terminal is divided into N × N small blocks by a block circuit 2 and orthogonally transformed by an orthogonal transformation circuit 3. Is done. The orthogonal transform coefficient output from the orthogonal transform circuit 3 is quantized by a quantization circuit 4 (a selected quantizer) having a plurality of quantizers, and the quantized value is sent to an output terminal 5. On the other hand, in the edge area detection circuit 6, the edge detection of the image signal from the input terminal 1 is performed using an edge detection operator typified by LaBrazian. The flat portion detection circuit 7 determines whether a certain pixel of the image signal from the input terminal 1 is a flat region. The block determination circuit 8 determines from the output of the edge region detection circuit 6 and the output of the flat portion detection circuit 7 whether or not the block is an edge region / flat region mixed block. The determination result is sent to the receiving side (output terminal 5) together with the quantization coefficient value.

The quantization circuit 4 selects a quantizer to be applied to the orthogonal transform coefficient from the orthogonal transform circuit 3 based on the determination result of the block determination circuit 8. That is, when the entire block is flat, or when the entire block has a complicated structure, the above-described noise does not occur or it is difficult to detect even if it occurs. And a quantizer that narrows the quantization of the orthogonal transform coefficient in order to prevent noise from being generated in the flat portion in a block in which edges and flat portions are mixed.
As a result, information can be efficiently distributed.

It is clear that this method can be applied to a case where an orthogonal transformation is performed on an inter-frame difference signal between a pixel value of one frame before and a pixel value of the current frame.

FIG. 2 shows a second embodiment of the present invention. Reference numeral 31 denotes a discrete cosine transform circuit as orthogonal transform means, which uses an efficient discrete coin transform as the orthogonal transform. Reference numeral 9 denotes a high-pass filter for inputting image means from the input terminal 1. In FIG. 3, the filter output value x 'of the pixel x is given by x' = 4x- (a + b + c + d). The threshold value Th1 set by the threshold value setting circuit 10 for x 'from the filter 9 is compared in the comparator 11, and as the comparator output _BH , when x'≥Th1, _BH = 1 x'<Th1 , B _H = 0 is obtained. In general, B _H = 1 when a certain pixel is a pixel in the edge area. On the other hand, in the low-pass filter 12 for inputting the image signal from the input terminal 1, the filter output x ″ of the pixel x is output.
X ″ = (4x + a + b + c + d) / 8, and the value x before filtering in the subtractor 13 is calculated.
Is calculated.

Δx = xx ″ The absolute value circuit 14 to which the output of the subtractor 13 is input has the absolute value | Δ
After x | is taken, the threshold value Th2 set by the threshold value setting circuit 15 is compared in the comparator 16, and as the comparator output B _L , when | Δx | <Th2, B _L = 1 | Δx | When ≧ Th2, B _L = 0 is obtained. Generally, when a certain pixel is a pixel in a flat area
B _L = 1.

A portion 100 surrounded by a dotted line is a detection unit for a mixed block of an edge area and a flat area, and the counting circuit 17 counts the number _NL of pixels of B _L = 1 in the block based on the output of the comparator 16. The N _L from the counter circuit 17 compares in a comparator 19 with a threshold value Th3 which is set by the threshold setting circuit 18, as a comparator output _{B N, B N = 1 N} L when N _L ≦ Th3 < When Th3, B _N = 0 is obtained. Generally, when there are many flat parts in a block, B _N =
It becomes 1. In the block determination circuit 20, if there is at least one pixel of B _H = 1 in the block of B _N = 1, the block is regarded as an edge area / flat area mixed block, and a discrete cosine is applied to the quantization circuit 4. As the quantizer for the orthogonal transform count of the output of the transform circuit 31, one having a small dead zone and a small step size is selected, and if not, a coarse one is selected.

FIG. 4 shows a third embodiment of the present invention. In this example,
The mixed block of the edge area / flat area is detected as follows.

First, in the sub-blocking circuit 21, the image from the input terminal 1 is divided into M × M (M: integer) small blocks. This small block is selected to be a sub-block of the block used in the orthogonal transform as shown in FIG. For example, if N = 8 and M = 4, the orthogonal transform block is the fifth
As shown in the figure, it is divided into four sub-blocks. Generally, the number S of sub-blocks is S = (N / M) ² . A variance value is calculated by a variance calculation circuit 22 for each subblock in one orthogonal transform block, and each variance value σ ² i (i = 1, 2,..., S) is set by a threshold circuit 23 The threshold value Ths is compared by the comparator 24. The comparator output C is determined by Ci = 1: σi ² ≦ Ths Ci = 0: σi ² <Ths (i = 1, 2,..., S). In general, when an edge component is present in a sub-block, the variance value is large, so that Ci = 1, and when the sub-block is flat, Ci = 0. Using the Ci, the block determination circuit 25 sets a parameter C for determining whether the orthogonal transformation block to be quantized is an edge area / flat area mixed block as follows.

C = C ₁ C ₂ C ₃ ... Cs Here, it is an exclusive OR. That is, C = 0 only when all of C _{1 to} Cs are 1 or 0, and otherwise C = 1. When C = 1, a sub-block having an edge component and a flat sub-block are present in the block at the same time, and thus can be regarded as an edge area / flat area mixed block. Therefore, in the block determination circuit 25, if C = 1, the block is regarded as an edge area / flat area mixed block, and the quantization circuit 4 determines the dead zone as a quantizer of the orthogonal transform coefficient of the corresponding block. By performing control such that a fine step size is selected and a coarse step size is selected otherwise, an error diffused to a flat region around an edge can be reduced.

[Effects of the Invention] As described above, according to the present invention, it is possible to efficiently remove noise generated in a flat area when a block includes a considerably strong edge and a flat part in orthogonal transform coding. And the image quality of the decoded image can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention, FIG. 2 is a block diagram of a second embodiment of the present invention, and FIG. 3 is a diagram showing pixels used in a high-pass filter and a low-pass filter. FIG. 4 is a block diagram of a third embodiment of the present invention, and FIG. 5 is a diagram showing a relationship between orthogonal transform blocks and sub-blocks. DESCRIPTION OF SYMBOLS 1 ... Input terminal, 2 ... Blocking circuit, 3 ... Orthogonal transformation circuit, 4 ... Quantization circuit, 5 ... Output terminal, 6 ... Edge area detection circuit, 7 ... Flat part detection circuit, 8 …… Block decision circuit.

Claims (1)

(57) [Claims] 1. A dividing means for dividing an image signal into N × N (N = integer) blocks, and an orthogonal transform for performing a discrete orthogonal transform for each block signal obtained by the dividing means to obtain a transform coefficient. Means, quantizing means for quantizing transform coefficients obtained by the orthogonal transform means so that a quantization interval can be changed, and at least a plurality of sub-blocks in the block for the signal of each block of the image signal. Detecting means for detecting whether the sub-block is an edge area or a flat area for each of the sub-blocks;
Edge area / flat area mixed block extracting means for extracting a block in which one or more edge area sub-blocks are mixed with flat area sub-blocks, and the edge area and the flat area are mixed based on the extraction result of the extracting means Means for controlling a quantization interval of said quantization means so as to quantize a transform coefficient of a block more finely than other blocks.