JP3305480B2 - Image encoding / decoding 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 encoding apparatus for compressing, encoding and transmitting an image signal by orthogonal transformation, and a decoding apparatus for decoding a transmitted code and reproducing an original image signal.

[0002]

2. Description of the Related Art In general, since a video signal has a very large amount of information, when recording or transmitting, a method of reducing the amount of information by high-efficiency coding so that image quality deterioration is not visually noticeable is known. Used. One of the high-efficiency codings is orthogonal transform coding. In the orthogonal transform coding, a video signal is divided into blocks, frequency decomposition is performed for each block by using orthogonal transform, information is compressed in consideration of visual characteristics, and coding is performed.

FIG. 5 is a diagram showing a configuration of a conventional image encoding / decoding device. In the figure, (a) is an encoding device, (b)
Indicates the configuration of the decoding device. First, the encoding device will be described. The image signal input from the input terminal 100 is divided by the blocking unit 101 into rectangular blocks of n × n pixels. Discrete cosine transform (hereinafter referred to as DCT) unit 1
02 performs DCT for each of the divided blocks to generate DCT coefficients. The quantization unit 103 quantizes the DCT coefficient by comparing it with a quantization table to generate a quantization coefficient. The coding unit 104 performs variable length coding on the quantized coefficients with reference to the Huffman code table. The encoded information is sent to the decoding device via the transmission path.

Next, the structure of the decoding apparatus will be described. The decoding unit 111 decodes the transmitted information into a quantized coefficient with reference to the code table. The quantized coefficient is inversely quantized by an inverse quantization unit 112 and inverse DCT by an inverse DCT unit 113.
Is returned to the original image signal. Inverse block synthesis unit 11
In step 4, a reproduced image is generated by combining the image signals restored for each block. However, in the above-described encoding / decoding device, the image signal is divided into rectangular blocks of n × n pixels regardless of the content thereof. Noise unique to orthogonal transform coding, such as mosquito noise due to the effect of image formation, has occurred, causing image quality degradation.

On the other hand, in recent years, there has been proposed an area division coding method in which a block is divided in accordance with the content of an image, and a block shape and a pixel value in the block are separated and coded. Have been reported. In addition, a technique in which orthogonal transform coding is combined with area division coding can achieve high coding efficiency and excellent image quality (for example, IEICE Transactions BI, No. 5, p. Variable Block Shape KL Transform Coding ”). In this method, an image is divided into a plurality of rectangular blocks, the rectangular blocks are deformed in accordance with the content of the image, and orthogonal transformation is performed for each block in accordance with the shape to encode the blocks.

[0006]

However, in the above method, the shape of the rectangular block is various, and the orthogonal transformation processing according to the block shape is more complicated than the conventional two-dimensional DCT or the like. Therefore, there is a problem that the encoding / decoding process takes too much time.

[0007]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has been made to reduce the processing time of orthogonal transform and inverse orthogonal transform according to a block shape, and is also applicable to moving picture coding. An object of the present invention is to provide a usable image encoding / decoding device.

[0008] A first invention is a blocking means for dividing an image signal into a plurality of polygon blocks composed of a plurality of pixels,
A block shape deforming unit for deforming the shape of the polygon block according to the content of the image, and a polygon block deformed by the polygon block deformed by the block shape deforming unit in a horizontal direction with respect to pixels in each row. Shape conversion means for performing interpolation processing and performing vertical interpolation processing on pixels in each column to convert the blocks into n × n pixel blocks, and orthogonal transformation means for performing orthogonal transformation for each of the shape-converted blocks Quantization means for quantizing the transform coefficients generated by the orthogonal transform means to generate the quantized coefficients; and coding means for coding the quantized coefficients and the shape information output from the block shape deforming means. This is an image coding device composed of: A second aspect of the present invention is an image encoding apparatus according to the first aspect, wherein in the interpolation processing, a value of a reference pixel having a short distance from the two reference pixels to the interpolation target pixel is inserted in order to obtain an interpolation target pixel value. Third
The invention of the first aspect is an image encoding apparatus according to the first aspect, wherein the interpolation processing inserts an average value of pixel values of two reference pixels in order to obtain an interpolation target pixel value. A fourth invention is an image coding apparatus according to the first invention, wherein in the interpolation processing, a value obtained according to a distance to a pixel value of two reference pixels is inserted in order to obtain a pixel value to be interpolated.

According to a fifth aspect of the present invention, there is provided decoding means for restoring a quantized coefficient and shape information of a polygon block from a transmitted code, and a shape parameter for restoring a polygon block from the shape information. A shape parameter generating means for generating, a dequantizing means for dequantizing the quantized coefficient to generate a transform coefficient, an inverse orthogonal transform means for performing an inverse orthogonal transform on the transform coefficient, and an output of the inverse orthogonal transform means. Some nx
inverse shape conversion means for performing interpolation processing on a block of n pixels in the vertical and horizontal directions based on the shape parameters to convert the block into polygon blocks, and block synthesis means for synthesizing the polygon blocks and restoring an image signal An image restoration device comprising:

[0010]

According to the image encoding / decoding apparatus of the present invention, various block shapes are converted into rectangular blocks of nxn pixels and converted using DCT, which is a fast algorithm proposed. -The decoding processing time can be reduced. Therefore, the present invention can be applied not only to an image encoding / decoding device that handles still images but also to a moving image encoding / decoding device.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an image encoding / decoding device according to the present invention will be described below in detail with reference to the drawings. FIG. 1 is a block diagram showing an apparatus configuration of an embodiment of an image encoding apparatus according to the present invention. The image signal input to the input terminal 10 is sent to the blocking unit 11. The blocking unit 11 sets vertices for dividing the image signal into a plurality of polygon blocks including a plurality of pixels. Reference numeral 12 denotes a block shape deforming unit that deforms the shape of the polygon block by moving the vertices according to the content of the image. The output of the block shape transforming unit 12 is output from the shape converting unit 13 and the encoding unit 1.
Sent to 6. The shape conversion unit 13 converts the shape of the polygon block deformed by the processing described later into a rectangular block of 8 × 8 pixels. The DCT unit 14 performs DCT on the rectangular block to generate DCT coefficients. Quantization unit 1
5 quantizes the DCT coefficient with reference to a quantization table. The encoding unit 16 includes the block shape deforming unit 12
And the quantization coefficient from the quantization unit 15 are encoded with reference to the code table, and sent out to the transmission path.

Next, the processing of the encoding apparatus according to this embodiment will be described. The image signal input to the input terminal 10 is stored in the blocking unit 11. The blocking unit 11 sets vertices so as to divide the input image signal into a plurality of rectangular blocks including a plurality of pixels. The position of the vertex in the image can be counted in units of pixels, and the coordinates of the vertex can be designated by a set of numbers in the horizontal and vertical directions. In this way, the coordinates of the vertices forming the quadrangle are stored in an array so that they can be called up by unique serial numbers. A square block is represented by four sets of serial numbers, and the square block itself is stored in an array so that it can be called with a unique serial number. Next, in the block shape deformation unit 12, the square block is deformed according to the image. The deformation of the rectangular block is realized by moving the vertices constituting the rectangular block.

FIG. 2 is a diagram showing how the vertices move.
Evaluation parameters Sa, Sb, Sc, and Sd are calculated for four rectangular blocks a212, b214, c216, and d218 that share the moving target vertex 210. The sum of Sa to Sd is used as the evaluation value T0 of the initial position. Here, the evaluation parameter is, for example, a sum of squares of AC components of pixel values in the square block (a value obtained by subtracting an average value in the block from each pixel value). Next, the evaluation parameters are also calculated for the four rectangular blocks 222, 224, 226, and 228 when the target vertex 210 is moved to the state of 220, and the evaluation value T1 as the sum thereof is obtained. The same processing is performed for n around the initial position of the target vertex 210.
The evaluation is performed on the number of movement candidate points, the evaluation values T0 to Tn are obtained, and the vertex 210 is moved to the position showing the minimum value. The above processing is performed in the order of the vertex-specific serial numbers. The above-described processing is repeated a plurality of times for all vertices, and the vertices are moved according to the image. The new coordinates of the vertex moved in the above process are sent to the encoding unit 16 as shape information. Instead of the coordinates of the vertices, the amount of displacement of each vertex from the initial position may be sent to the encoding unit 16 as shape information.

FIGS. 3 and 8 are diagrams for explaining the processing in the shape conversion unit 13. FIG. The square block 300 includes four vertices 310, 312, 314, and 316, and converts the square block 399 into a rectangular block of 8 × 8 pixels. In this conversion processing, first, the block is operated in the horizontal direction line by line, and the k-th row (k = 0, 1,...,
The pixel value of (K-1); K is the number of rows in the square block 300) is extracted into the one-dimensional array 320, and the number m of data is counted. Next, an array 320 having m data
Is converted into eight data while interpolating the data.
Now, m data are expressed as {x (0), x (1),.
-1)}, the eight pieces of data X (i) (i =
0, 1,..., 7) are as follows. X (0) to X
Assuming that the distance to (7) and the distance from x (0) to x (n-1) are 1, the data string of data X (i) {x
(0), x (1),..., X (n−1)}, corresponding point x ′
(I) (i = 0, 1,..., 7) is x (d) (d = 0,
1,..., M−2) and x (d + 1). Therefore, the value of x '(i) can be obtained from x (d) and x (d + 1). For example, the value of x (d) and x (d + 1) that is closer to x ′ (i) is x ′ (i). As another method, x (d) and x (d +
The average value of 1) is calculated and x '(i) or x' (i)
It is also effective to interpolate and obtain the data according to the distance to (i).

Next, the two-dimensional array 330 is scanned in the vertical direction, the data in the j-th column (j = 0, 1,..., 7) is extracted into the one-dimensional array 320, and the number of data is counted. . Subsequently, the array 320 having K data is expanded to eight data while interpolating the data, and is stored in the j-th column of the second two-dimensional array. The above processing is repeated for all the columns in the rectangular block. When the above processing is completed, the square block 300 is converted into an 8 × 8 rectangular block on the second two-dimensional array.

The rectangular block data is stored in the following DCT unit 1
4 and subjected to two-dimensional DCT.
5 and is sent to the encoding unit 16. The encoding unit 16 encodes the quantized coefficients and the shape information with reference to a code table prepared in advance, and sends the encoded information to a transmission path.

FIG. 4 is a diagram showing an apparatus configuration of an embodiment of an image decoding apparatus according to the present invention. In the figure, reference numeral 20 denotes a decoding unit that decodes a transmitted code and obtains a quantization coefficient and shape information. Reference numeral 21 denotes an inverse quantization unit that inversely quantizes the quantized coefficient to generate a transform coefficient. Reference numeral 22 denotes an inverse DCT unit that performs inverse DCT on the transform coefficients to restore rectangular block data of 8 × 8 pixels. Reference numeral 25 denotes a shape parameter generation unit which generates shape parameters for block shape restoration from the vertex coordinates as the shape information by a process described later. Reference numeral 23 denotes an inverse shape conversion unit which restores a square block from the rectangular block and the shape parameters. The block combining unit 24 combines the restored rectangular blocks to generate a reproduced image.

Next, the processing of the decoding apparatus according to the present embodiment will be described. The transmitted code is decoded by the decoding unit 20 into quantization coefficients and shape information while referring to the same code table as the coding unit 16, and is supplied to the inverse quantization unit 21 and the shape parameter generation unit 25, respectively. You. The quantized coefficients are inversely quantized in the inverse quantization unit 21 with reference to the same quantization table as the quantization unit 15, and the inverse DCT
An inverse DCT is performed by the unit 22 to restore a rectangular block of 8 × 8 pixels. On the other hand, the shape parameter generation unit 25 obtains shape parameters for restoring the original square block shape from the decoded vertex coordinates. First, the vertical coordinates of the four vertices are compared, and the minimum value ymin and the maximum value ymin are compared.
max is obtained, and the number of rows K (K = y) in the original square block is obtained.
max-ymin). Next, a rectangular block composed of four vertices is sequentially scanned in the horizontal direction, and the number of data n (k) (k = 0, 1,..., (K−
1)) is checked and stored in association with the line number. The number of rows K of the rectangular block and the number of data n (k) of each row obtained by the above processing are sent to the inverse shape conversion unit 23 as shape parameters together with the coordinates of the vertices. When the processing is started, the inverse shape conversion unit 23 first scans the rectangular blocks in the vertical direction in order, and the p-th column (p = 0, 1,..., 7)
Are extracted into a one-dimensional array. Subsequently, the number of data is expanded by data interpolation so that the eight pieces of data in the one-dimensional array become K pieces of data, and stored in the p-th column of the two-dimensional array. At the end of the process up to the seventh column, the 8 × 8 rectangular block becomes an 8 × K rectangular block on the two-dimensional array.

Next, the rectangular blocks are sequentially scanned in the horizontal direction, and the q-th row (q = 0, 1,..., (K−
1)) is extracted into a one-dimensional array. Subsequently, the number of data is expanded so that the eight data in the one-dimensional array becomes the number of data n (q) in the corresponding row. By performing the above processing for all rows, a square block is restored. The block synthesis unit 24 pastes the restored square block at a position corresponding to the vertex coordinates,
Generate a playback image.

Next, an encoding apparatus according to a second embodiment of the present invention will be described. The encoding device of the second embodiment is the same as the encoding device of FIG.
It can be realized with the same configuration as the encoding device of the embodiment. The image signal passes through the input terminal 10 and is stored in the blocking unit 11. The blocking unit 11 sets vertices so as to divide the input image into a plurality of triangular blocks including a plurality of pixels. The position of the vertex in the image can be counted in units of pixels, and the coordinates of the vertex can be designated by a set of numbers in the horizontal and vertical directions. In this way, the coordinates of the vertices constituting the triangular block are stored in an array so that they can be called up by unique serial numbers. The triangular block is represented by the three sets of serial numbers, and the triangular block itself is stored in an array so that it can be called with a unique serial number. Next, in the block shape deformation unit 12, the triangle block is deformed according to the image. The transformation of the triangle block is realized by moving the vertices constituting the triangle block.

FIG. 6 is a diagram showing how the vertices move.
A hexagonal area 600 composed of six triangular blocks sharing a moving target vertex 610 is set as a target area, and evaluation parameters Sa are set for the triangle blocks a to f.
〜Sf is calculated. The sum of Sa to Sf is used as the evaluation value T0 of the initial position. Here, the evaluation parameter is
As in the case of the square block, the square sum of the AC components of the pixel values is used.

Next, evaluation parameters are also calculated for the six triangular blocks a 'to f' when the target vertex 610 is moved to the state of 620, and an evaluation value T1 as a sum thereof is obtained. The same process is performed on n moving candidate points around the initial position of the target vertex 610 to obtain evaluation values T0 to Tn, and the vertex 610 is moved to the position showing the minimum value. The above processing is performed in the order of the vertex-specific serial numbers. The process described above is repeated a plurality of times for all vertices,
The vertices are moved according to the image to deform the triangular block. The vertex coordinates of the transformed triangular block are sent to the encoding unit 16 as shape information, and are encoded. The deformed triangle block is then converted by the shape conversion unit 13 into a rectangular block of 8 × 8 pixels as in the case of a square block. FIG. 7 is a diagram illustrating processing in the shape conversion unit 13 according to the second embodiment. The shape conversion unit 13
The inside of the triangular block 700 composed of the vertices 702, 704, and 706 is sequentially scanned in the horizontal direction, the pixel values of each row are extracted into a one-dimensional array, and the data is interpolated to increase the number of data to eight. When the scanning in the horizontal direction is completed, the converted rectangular blocks are sequentially scanned in the vertical direction, and the number of data of the extracted one-dimensional data string is enlarged or reduced to eight. Thus, the triangular block is converted into a rectangular block of 8 × 8 pixels. The rectangular block is DC
After the DCT is performed in the T unit 14, it is quantized in the quantization unit 15, encoded in the encoding unit 16, and sent to the transmission path together with the encoded shape information.

Next, a decoding apparatus according to a second embodiment of the present invention will be described. The device configuration of the decoding device of the second embodiment is also
This can be realized with the same device configuration as the decoding device of the first embodiment shown in FIG. The shape information decoded by the decoding unit 20 is sent to the shape parameter generation unit 25, and generates shape parameters for restoring the original triangle block from the coordinates of the vertices as the shape information. That is, the number of rows in the triangular block and the number of data in each row are obtained by the same processing as that of the shape parameter generation unit of the first embodiment, and are sent to the inverse shape conversion unit 23 as the shape parameters together with the coordinates of the three vertices. The inverse shape converter 23 converts the 8 × 8 pixel rectangular block data, which has been inversely quantized and subjected to inverse DCT, based on the shape parameters into the original triangular block in the same manner as in the first embodiment. . The restored triangular block is pasted at a position corresponding to the vertex coordinates in the block combining unit 24, and a reproduced image is generated.

[0024]

According to the image encoding / decoding apparatus of the present invention, the encoding / decoding is performed by performing block division according to the content of the image.
Since decoding is performed, block distortion and mosquito noise can be reduced, and the image quality of a reproduced image is also improved. In addition, a block shape deformed so as to increase the conversion efficiency is converted into an n × n rectangular block, and a high-speed algorithm D is proposed.
Since the orthogonal transform is performed using the CT, not only the conversion efficiency is improved, but also the processing time for encoding and decoding can be shortened. Therefore, the present invention can be applied to not only a still image but also a moving image encoding / decoding device.

[Brief description of the drawings]

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

FIG. 2 is a view for explaining processing of a block shape deforming means according to the first embodiment of the present invention.

FIG. 3 is a view for explaining processing of a shape conversion unit according to the first embodiment of the present invention.

FIG. 4 is a block diagram illustrating a device configuration of an image decoding device according to the present invention.

5A is a diagram illustrating a configuration of a conventional image coding, and FIG. 5B is a diagram illustrating a configuration of a decoding device.

FIG. 6 is a view for explaining processing of a block shape deforming means according to a second embodiment of the present invention.

FIG. 7 is a view for explaining processing of a shape converting means according to the second embodiment of the present invention.

FIG. 8 is a view for explaining the processing of the shape conversion means of FIG. 3;

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 Input terminal 11 Blocking part 12 Block shape deformation part 13 Shape conversion part 14 DCT part 15 Quantization part 16 Encoding part 20 Decoding part 21 Inverse quantization part 22 Inverse DCT part 23 Inverse shape conversion part 24 Block synthesis part 25 shape parameter generation unit 101 blocking unit 102 DCT unit 103 quantization unit 104 encoding unit 111 decoding unit 112 inverse quantization unit 113 inverse DCT unit 114 block synthesis unit 210, 220 vertices 212, 214, 216, 218 square block 222, 224, 226, 228, 300 square block 310, 312, 314, 316 vertex 320 one-dimensional array 330 two-dimensional array 600 hexagonal region 610, 620 vertex 700 triangular block 702, 704, 706 vertex

Continuation of the front page (56) References JP-A-3-89792 (JP, A) JP-A-5-316371 (JP, A) Ichiro Matsuda, Susumu Ito, Toshio Utsunomiya, Adaptive Variable Block KL Transform Coding of Images , IEICE Transactions B-1, Japan, IEICE, Vol. J76-BI, No. 5, pp. 399-408 Michael GILGE, Tho mas ENGELHARDT, Ral f MEHLAN, CODING OF ARBITRARILY SHAPE D IMAGE SEGMENTS B ASED ON A GENERALI ZED ORTHOGONAL TRA NSFORM, Signal Proc essing: Image Commu nication, Elsevier Science Publishers B. V. , 1, pp. 153−180 (58) Fields surveyed (Int. Cl. ⁷ , DB name) H04N ^7/ 24-7/68

Claims (5)

(57) [Claims] 1. Blocking means for dividing an image signal into a plurality of polygon blocks composed of a plurality of pixels; block shape deformation means for deforming the shape of the polygon block according to the content of an image; Each polygon block deformed by the polygon block deformed by the deformation means
Perform horizontal interpolation on the pixels in the row and
By performing vertical interpolation on the pixels in a column,
And shaping means for converting a block of n × n pixels Te, and orthogonal transformation means for performing shape transformed orthogonal transformation for each of the blocks, the quantized coefficients by quantizing the transform coefficients generated by said orthogonal transformation means An image coding apparatus comprising: a quantizing means for generating; and a coding means for coding the quantization coefficient and shape information output from the block shape deforming means. 2. In the interpolation processing, a pixel value to be interpolated is obtained.
Distance to the interpolation target pixel of the two reference pixels
Wherein a value of a reference pixel having a value close to is inserted.
2. The image encoding device according to 1. 3. In the interpolation processing, a pixel value to be interpolated is obtained.
The average of the pixel values of the two reference pixels
2. The image encoding apparatus according to claim 1, wherein: 4. In the interpolation processing, a pixel value to be interpolated is obtained.
To the pixel values of the two reference pixels.
2. The method according to claim 1, wherein the determined value is inserted.
Image coding device. 5. A decoding means for restoring a quantized coefficient and shape information of a polygon block from a transmitted code, and a shape parameter generation for generating a shape parameter for restoring a polygon block from the shape information. Means, inverse quantization means for inversely quantizing the quantized coefficients to generate transform coefficients, inverse orthogonal transform means for inverse orthogonal transforming the transform coefficients, and n × n pixels which are outputs of the inverse orthogonal transform means. Vertical and water based on the shape parameters
An image decoding apparatus, comprising: an inverse shape conversion unit that performs interpolation processing in a horizontal direction to convert into a polygon block; and a block synthesis unit that synthesizes the polygon block and restores an image signal.