JPH06339030A - Area shape encoding and decoding system - Google Patents

Abstract

PURPOSE:To provide an encoding and decoding system in which an area shape with little deterioration can be encoded and decoded by vector quantization by the small number of storage shape patterns. CONSTITUTION:The plural shape patterns are preliminarily stored in a shape memory 101 at an encoding side, and in a shape memory 201 at an encoding side. At the time of encoding the shape, a shape pattern to be encoded is inputted, the stored shape is successively taken out from a shape memory 101, a deforming operation is operated by a shape deforming part 103, and a reference shape is prepared. Next, the matching of the input shape with the reference shape is obtained by a shape matching part 102, and similarity is calculated. Then, the shape pattern whose similarity is the highest and a deformation parameter are selected and outputted. At the decoding side, the identification number of the shape is received, the shape is selected from a shape memory 201, the deformation parameter is received, the deforming operation indicated by the deformation parameter is operated by a shape deforming part 203, and the obtained shape is outputted as the decoded shape.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for coding / decoding a region shape in image coding / decoding.

[0002]

2. Description of the Related Art In the area information compression used in the "image data compression method" described in JP-A-63-164573, an image is divided into square blocks, and each block is divided into two areas. The division pattern is vector-quantized. That is, two fixed-size divided blocks
Divide by the representative color of the colors, represent the division pattern as the binarized data, and select the division pattern having the smallest distortion with respect to the division pattern to be encoded from the division patterns given in advance in the code table.

[0003]

In the above-mentioned conventional technique, one area is divided into a plurality of blocks, and a label for identifying the area must be given to the divided areas, which reduces efficiency. Further, one block cannot be divided into three or more areas, the size of the division pattern and the matching direction are constant, and there is no means for separately encoding even if the similarity due to matching is small. Depending on the type, there is a drawback that the mismatch with the accumulation pattern increases and deterioration occurs. In order to prevent this, it is conceivable to increase the number of storage patterns, but a large storage capacity for storing the shape patterns is required.

It is an object of the present invention to provide an encoding / decoding method by vector quantization of a region shape with a small number of accumulated shape patterns and little deterioration.

[0005]

According to a first aspect of the present invention, in an arbitrary area shape encoding / decoding method cut out from an image, the encoding side stores means for accumulating a plurality of shape patterns. A means for deforming the shape by rotating, turning over, enlarging, reducing, and matching the area shape to be encoded with the deformed shape, select the shape and deformation parameter with the highest degree of similarity, and select the identification number of the selected shape and And a means for outputting a transformation parameter, wherein the decoding side
A means for accumulating a plurality of shape patterns, a means for extracting the shape pattern from the accumulating means by the shape identification number, and a means for deforming the shape pattern by the deformation parameter to generate a region shape. .

A second aspect of the invention is an arbitrary region shape encoding / decoding method cut out from an image, wherein the encoding side is
A means for accumulating a plurality of shape patterns, a means for deforming the accumulated shapes by rotating, turning over, enlarging, reducing, and matching the area shape to be encoded with the deformed shape, A deformation parameter is selected, the identification number of the selected shape and the deformation parameter are output, a threshold value for the similarity is set in advance, and if there is no shape with a similarity degree equal to or higher than the threshold value in the shape memory, the shape is stored. A means for separately encoding and outputting a shape code without selecting from the means, and outputting a code for separately shape-encoding as an identification number of the shape, and the decoding side has means for accumulating a plurality of shape patterns, Means for extracting a shape pattern from the storage means by the shape identification number; means for deforming the shape pattern by the deformation parameter to generate a region shape If it said shape identification number is a code that is separately shape coding, the coded shape decoding the shape from, and means for outputting the decoded shape, characterized in that.

[0007]

According to the present invention, a plurality of shape patterns are stored, and the region shape to be encoded is selected as the deformed shape having the highest similarity as compared with the shape obtained by performing a specific deformation operation on the stored shape. Then, the identification number and the deformation parameter of the selected shape are encoded. If necessary, if a shape with a high degree of similarity cannot be found in the shape pattern storage means, it is encoded separately.

As described above, since the area to be coded is not divided and the area having an arbitrary size is coded with one group of codes, the coding efficiency per area can be increased.

Further, since the shapes accumulated in the shape pattern accumulating means are transformed and matching is performed, it is possible to increase the kinds of shapes that can be expressed even with the same memory amount, and it is possible to relatively reduce the number of accumulated shapes. . Further, by performing a transforming operation and introducing a separate encoding means, not only the accumulated shape but also other shapes can be expressed, so that the encoding error can be reduced.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a diagram for explaining the area shape encoding / decoding method of the first invention. FIG. 1A shows an encoding unit on the transmission side, and the encoding unit is the shape memory 10
1, a shape matching unit 102, and a shape deforming unit 103. FIG. 1B shows a decoding unit on the receiving side, which includes a shape memory 201 and a shape transforming unit 203.
Composed of.

In this method, first, a plurality of shape patterns are stored in the shape memory 101 on the encoding side and the shape memory 201 on the decoding side in advance. The same data is stored in these two shape memories. As the accumulated data, the maximum values of the vertical and horizontal widths of the shape are determined, and the vertical and horizontal widths of the shape data are enlarged / reduced so that they match the specified maximum values, and standardized. Register Here, the method of deforming the shape will be described later.

When the shape is encoded, the shape pattern to be encoded is input, the accumulated shapes are sequentially fetched from the shape memory 101, and the shape deforming unit 103 performs rotation, turning, enlargement / reduction, etc. Deformation operation is performed to create a reference shape. The transforming operation expands / reduces the standardized and stored shapes so that the shapes can be generated by rotation or inside out so that the shapes match the maximum widths in the vertical and horizontal directions of the input shape.

Next, in the shape matching unit 102, the input shape and the reference shape are matched and the similarity is calculated. Then, the shape pattern and the deformation parameter having the highest similarity are selected, and the identification number and the deformation parameter of the selected shape are output. In this case, the deformation parameters are the rotation angle, the presence / absence of the flip operation, and the enlargement / reduction ratio. The method of calculating the degree of similarity will be described later.

On the decoding side, the shape identification number is received, the shape is selected from the shape memory 201, the transformation parameter is further received, and the transformation operation indicated by the transformation parameter is performed in the shape transformation unit 203 to decode the obtained shape. Output as a converted shape.

Next, a method of deforming the shape will be described. First, regarding rotation, 90 °, 180 °, 270 °
It can be easily realized by coordinate conversion by limiting to. The same applies to inside out. Since inside out is combined with 180 °, either inside out for the horizontal direction or inside out for the vertical direction may be prepared.

Next, the enlarging / reducing method will be described. Regarding the representation of the area shape, the maximum width m in the horizontal direction and the maximum width n in the vertical direction are obtained, and it is assumed that the inside of the area is represented by 1 and the outside of the area is represented by 0 in an m × n matrix. Consider, for example, transforming the area shape of m × n pixels to the size of p × q pixels. First, the greatest common divisor of the size before and after deformation in each direction is obtained. here,
A = L. C. M. (M, p) B = L. C. M. (N,
q). Next, the area shape of m × n is A / m in the lateral direction.
2 times, and B / n times in the vertical direction to form an intermediate region shape of A × B pixels. One pixel of the original pixel is A /
Corresponding to a block of m + B / n, all pixel values in the block inherit the original pixel value of the corresponding position. Next, this intermediate region shape is divided into A / p pixels in the horizontal direction and B / q pixels in the vertical direction to form p × q blocks of A / p × B / q pixels. This one block corresponds to one pixel of the modified area shape. As for the method of determining each pixel value, the pixel value is determined to be the one with the most "0" or "1" by the majority function in the corresponding block. Here, (m, n, p, q) can be used as the deformation parameter for enlargement / reduction. However, if the size of the accumulated shape is standardized, the size before deformation is constant, so that the deformation parameter is only (p, q).

A method of calculating the similarity will be described. The degree of similarity is defined between the input area shape to be encoded and the reference area shape extracted from the shape memory and subjected to the deformation operation. As the simplest degree of similarity, when the area shape expression is defined as 1 inside the area and 0 outside the area as described above, the exclusive OR of the input area shape and the reference area shape (X
OR), and the number of pixels that become 1 (that is, the number of mismatched pixels) is counted, and this can be defined as the degree of similarity.

FIG. 2 is a diagram for explaining the area shape encoding / decoding method of the second invention. FIG. 1A shows a coding unit on the transmission side, and the coding unit is the shape memory 30.
1, a shape matching unit 302, a shape deforming unit 303, a shape encoding unit 304, and a shape normalizing unit 305.
FIG. 2B shows the decoding unit on the receiving side, and the decoding unit includes a shape memory 401, a shape transformation unit 403, a shape decoding unit 404, and a shape normalization unit 405.

In this method, first, a plurality of shape patterns are stored in the shape memory 301 on the encoding side and the shape memory 401 on the decoding side in advance. The same data is stored in these two shape memories. As the accumulated data, the maximum values of the vertical and horizontal widths of the shape are set, and the vertical and horizontal widths of the shape data are enlarged / reduced to match the specified maximum values, standardized and registered. I'll do it. Further, a threshold value is set for the similarity obtained by the shape matching unit 302. The shape modification method and the similarity calculation method are the same as those in the above-described first invention.

When the shape is encoded, the shape pattern to be encoded is input, the accumulated shapes are sequentially fetched from the shape memory 301, and the shape deforming unit 303 performs rotation, turning, enlargement / reduction, etc. Deformation operation is performed to create a reference shape. The transforming operation expands / reduces the standardized and stored shapes so that the shapes can be generated by rotation or inside out so that the shapes match the maximum widths in the vertical and horizontal directions of the input shape.

Next, in the shape matching unit 302, the input shape and the reference shape are matched and the similarity is calculated. Then, the shape pattern and the deformation parameter with the highest similarity are selected. In this case, the deformation parameters are the rotation angle,
It is the presence / absence of the flip operation and the enlargement / reduction rate. Here, if the similarity of the matching result is larger than the specified threshold value, the identification number of the shape and the deformation parameter are output.
Otherwise, in the shape encoding unit 304, the shape of the input region shape is encoded by a conventional method by encoding the contour with a chain code or by polygon approximation to encode the vertex coordinates. It encodes and outputs the shape code.
Further, the coded shape is standardized by enlarging / reducing the coded shape to a specified size in the shape normalizing unit 305, and
, A new shape identification number is given, and this is used as a code indicating that it has been encoded separately, and is output.

Here, the coded shape does not necessarily need to be stored in the shape memory in view of the memory capacity and memory management. In this case, the identification number may be assigned a number that can identify that the shape memory has not been selected.

On the decoding side, the shape identification number is received, and if the shape indicated by the number is in the shape memory 401, that shape is selected, and the transformation parameter is further received.
The shape transformation unit 403 performs the transformation operation indicated by the transformation parameter, and outputs the obtained shape as a decoded shape.
Otherwise, it is determined that the shape coding is performed separately, the shape code is received, the shape decoding unit 404 decodes the shape, and outputs the decoded shape. Further, when the coding shape is stored in the shape memory 401 on the coding side, the decoding shape is also stored on the decoding side by the shape normalizing unit 4.
It is standardized in 05 and stored in the shape memory together with the received identification number.

[0024]

As described above, since the number of shape patterns to be stored can be reduced, the shape pattern recording capacity can be reduced. Further, since the coding error can be reduced, it is possible to realize a coding / decoding method with little deterioration by applying the method of the present invention to the image coding / decoding method.

[Brief description of drawings]

FIG. 1 is a diagram showing an encoding unit and a decoding unit of the first invention, respectively.

FIG. 2 is a diagram showing an encoding unit and a decoding unit of the second invention, respectively.

[Explanation of symbols]

 101, 201, 301, 401 shape memory 102, 302 shape matching unit 103, 203, 303, 403 shape transforming unit 304 shape encoding unit 404 shape decoding unit 305, 405 shape normalizing unit

Claims (2)

[Claims] 1. An arbitrary region shape encoding / decoding method cut out from an image, wherein the encoding side is means for accumulating a plurality of shape patterns, and the accumulated shape is rotated / inverted / enlarged / reduced. A means for deforming, a means for matching the area shape to be encoded and the deformed shape, selecting a shape and a deformation parameter with the highest degree of similarity, and outputting an identification number and a deformation parameter of the selected shape, The decoding side includes means for accumulating a plurality of shape patterns, means for extracting a shape pattern from the accumulating means by the shape identification number, and means for deforming the shape pattern by the deformation parameter to generate a region shape, A region shape encoding / decoding method characterized by the following. 2. An arbitrary region shape encoding / decoding method cut out from an image, wherein the encoding side has means for accumulating a plurality of shape patterns, and the accumulated shape is rotated / inverted / enlarged / reduced. Deformation means, matching the shape of the region to be encoded with the deformation shape, selecting the shape and deformation parameter with the highest similarity, outputting the identification number and deformation parameter of the selected shape, and If a threshold value is set and a shape having a degree of similarity equal to or greater than the threshold value is not in the shape memory, the shape is not selected from the accumulating means, the shape code is output separately, and the shape code is output separately. The decoding side has means for accumulating a plurality of shape patterns, and the shape side uses the shape identification number to accumulate the shape patterns. Means for extracting from the stage, means for deforming the shape pattern by the deformation parameter to generate a region shape, and when the shape identification number is a code that is separately shape-encoded, the shape is decoded from the shape code, A region shape encoding / decoding method, comprising: a means for outputting a decoded shape.