JPH10178638A - Character area encoding method, decoding method, character area encoder and decoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To drastically reduce the encoding information amount of a character part and to improve a compression rate of an image jumpingly. SOLUTION: First, a character area extracting part 102 extracts a character area 103 from an input image 101. Pixel values of character area 103 are vectorized in a character recognition part 104 and recognized as a character. As a result, a character code 105 is outputted. Various fonts 107 which correspond to the code 105 are retrieved from a font database 106 and inputted to a character matching part 108 together with an image of the area 103. The part 108 searches a font type 111, a character size 112 and a character color 113 which approximate the pixels of the area 103 most. A start position 109 of the character part, a character interval 110 and a character code 105 together with these data are outputted as encoded data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a method and an apparatus for encoding an image including characters.

[0002]

2. Description of the Related Art Conventionally, as a method of encoding an image including a character, after extracting an area including a character portion in the image, a PC
A method has been proposed in which encoding is performed using M or DPCM, and transform encoding such as DCT is performed on parts other than characters. For example, as shown in FIG. 5, there is a method in which conversion encoding is used for basic image processing, and only character portions are encoded by DPCM.

[0003] Transform coding is an effective technique for images that do not contain high frequency components, such as natural images, and is applied to regions that include steep signals that contain a large amount of high frequency components, such as characters. Then, due to the lack of high frequency components,
Block-like coding noise occurs. In order to avoid block-like coding noise, a large amount of information is required, and as a result, high compression cannot be achieved. Therefore, there has been proposed a method of applying a process such as PCM or DPCM that does not generate block-like coding noise even when there is a sharp signal change, such as transform coding, for a character portion.

[0004]

In the above-described conventional character area encoding method, an encoding method in pixel units such as PCM or DPCM is used. There was a drawback due to the increase, and the compression ratio was limited.

SUMMARY OF THE INVENTION It is an object of the present invention to greatly reduce the amount of encoded information in a character portion and to dramatically improve the image compression ratio.
An object of the present invention is to provide a character area encoding method and apparatus and a corresponding character area decoding method and apparatus.

[0006]

According to a first character area encoding method of the present invention, when encoding pixels of a character portion included in an image, the character portion is extracted as a character region and character recognition is performed. After that, character encoding is performed, a font closest to the pixel of the character portion is searched for among various character fonts corresponding to the character code, and parameters thereof and parameters necessary for reproducing the character portion are encoded. .

[0007] A first character area decoding method according to the present invention comprises:
A font corresponding to the character code is output, the two parameters are decoded, and pixels in the character area are reproduced and output.

[0008] When encoding pixels of a character portion included in an image, the character portion is extracted as a region and character recognition is performed. The recognition processing may be performed by an arbitrary method. For example, the method of Reference [1] can be used. Character recognition rate is 1
It need not be 00%. On both the encoding and decoding sides,
Various character vector fonts are stored as data line libraries.

The matching with the character portion is calculated by changing the size, color, and interval of various character vector fonts corresponding to the code for the recognized character. Any standard such as a mean square error, a mean absolute error, or a correlation at a pixel value level such as RGB or YCrCb can be used as a standard for this matching. The type, size and interval of the character font having the highest degree of approximation are found by matching.

The data to be encoded are the starting position of the character portion, the character code, the font type, the character size, the color and the interval, and if necessary, variable length encoding such as Huffman encoding or arithmetic encoding. And outputs it as compressed data.

The decoding side decodes the start position of the character portion, the character code, the font type, the character size, the color, and the character interval, and converts the font corresponding to the character code into one having the specified size and color. The character string is reproduced by arranging the character strings at the character intervals obtained from the start position of the character part in the image.

According to a second character area encoding method of the present invention, when encoding pixels of a character portion included in an image, the character portion is extracted as a character region, and after character recognition, the character encoding is performed. Performing a search for a font closest to the pixel of the character portion among various character fonts corresponding to the character code, encoding and outputting the parameters and parameters necessary for reproducing the character portion, and outputting the character code. Then, an approximate image of the character portion is generated from the character font corresponding to the above and both parameters, and is used as a predicted image for the input image, a prediction error is obtained, and the prediction error is encoded.

According to a second character area decoding method of the present invention,
A character area decoding method corresponding to the character area encoding method, wherein a font corresponding to the character code is output, the two parameters are decoded, and an approximate image of the character part is obtained from the font and the decoded parameters. Find the pixel value,
The prediction error is decoded, the pixel value of the approximate image is added to the decoded prediction error, and the pixel of the character area is reproduced and output.

The process up to the process of searching for the type of character font, character size and spacing by matching is the same as in the first method. Then, the starting position of the obtained character part,
Generates an approximate image of a character string determined by character code, font type, size, color, and character spacing. This approximate image is used as a predicted image for the input image, and a prediction error is obtained. The prediction error is calculated by predictive coding such as DPCM or DC.
The encoding is performed using a transform encoding such as T.

The encoding side encodes a start position of a character portion, a character code, a font type, a character size, a color and an interval, and encodes and outputs a prediction error.

On the decoding side, the starting position of the character portion, the character code, the font type, the character size, the color and the interval are decoded, and the font corresponding to the character code is restored to the one having the specified size and color. Then, the images are arranged at character intervals obtained from the start positions of the character portions in the image, and the decoded prediction errors are added to reproduce the image in the character region.

According to a first aspect of the present invention, there is provided a character region encoding device for extracting a character portion as a character region from an input image, a font database storing various fonts corresponding to character codes, Character recognition means for recognizing a character from a character area, outputting a character code, and searching the font database; and inputting a font output from the character area and the font database, and a font closest to a pixel of the character area. Exploring,
There is a character matching means for encoding and outputting the parameters and the parameters necessary for reproducing the character portion.

According to a second character area encoding apparatus of the present invention, a character area extracting means for extracting a character portion as a character area from an input image, a font database storing various fonts corresponding to character codes, Character recognition means for recognizing a character from a character area, outputting a character code, and searching the font database; and inputting a font output from the character area and the font database, and a font closest to a pixel of the character area. Exploring,
Inputting the parameters, a character matching means for encoding and outputting parameters necessary for reproducing the character portion, both parameters output from the character matching means, and a font output from the font database; There is an approximate character generating means for generating an approximate image of a portion, a difference means for obtaining a difference between the character area and the approximate image and outputting the difference as a prediction error, and a prediction error encoding means for encoding the prediction error.

The first character region encoding apparatus has a first
The character area decoding device stores a variety of fonts, outputs a font corresponding to the character code, a font database, decodes both parameters, and decodes both parameters and the pixels of the character area from the font. And a character generating means for reproducing and outputting.

The second character area encoding apparatus has a second
The character area decoding device stores a variety of fonts, outputs a font corresponding to the character code, a font database, decodes both parameters, and approximates the character portion from the decoded parameters and the font. Approximate character generation means for determining the pixel value of the image, prediction error decoding means for decoding the prediction error, and adding the pixel value of the approximate image and the decoded prediction error, to reproduce the pixels of the character area, It has addition means for outputting.

[0021]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of a character area coding apparatus according to a first embodiment of the present invention.

The character area encoding apparatus according to the present embodiment includes a character area extraction unit 102, a character recognition unit 104, a font database 106, and a character matching unit 108.

First, a character area 103 is extracted from an input image 101 by a character area extraction unit 102. In the pixel value of the character area 103, the character part is vectorized by the character recognition unit 104 and recognized as a character. As a result, the character code 105 is output. Character code 105
Are searched from the font database 106 and input to the character matching unit 108 together with the image of the character area 103.

The character matching unit 108 searches for a font type 111, a character size 112, and a character color 113 which best approximate the pixels of the character area 103 among the various fonts 107. Along with these data, the start position 109 of the character portion and the character interval 110 are output as encoded data.

FIG. 2 is a block diagram of a character area decoding apparatus corresponding to the character area encoding apparatus of FIG.

The character area decoding apparatus according to the present embodiment includes a character generation unit 201 and a font database 202.

Various fonts 203 corresponding to the character codes 105 sent from the character area encoding apparatus are searched from the font database 202, and the fonts 203 are coded data (character part data) also sent from the character area encoding apparatus. Start position 109, character spacing 110, font type 1
11, the character size 112, and the character color 113) are input to the character generation unit 201.

The character generator 201 calculates the pixel value of a character according to the following procedure. Selected character code 10
The font corresponding to the received font type 111 is selected from among the various fonts 203 corresponding to 5. For example, if the received character code 105 represents “A” and the font database 202 stores Mincho and Gothic fonts, the font “A” of Mincho and Gothic Sent.

Assuming that the received font type 111 is Mincho font and the received character size 112 is 20 points, the character generation unit 201 determines that the character shape is “Mincho font” including the line width. It is determined to be "A" of "20 points". The size and position of the character are determined by the start position 109 of the received character portion and the character interval 110.

The start position of a character may be a plurality of data. In this example, the case where the character size can be determined by the start position 109 of the character portion and the character interval 110 is shown. Must be received as encoded information.

The pixel value of the character is the character color 11 including the density.
Determined by 3. The information of the character color 113 may be received after being decomposed into only the color information and the only luminance information.

As described above, the shape, location, size, and color (brightness) of the character are calculated. If the sample points of the image correspond to the characters, the values can be calculated.

In this way, a font having a specified size and color is restored from the various fonts 203 corresponding to the character codes 105, and the character strings are arranged at the character intervals obtained from the start positions of the character portions in the image. It is reproduced and output as the pixel 204 of the character area.

FIG. 3 is a block diagram of a character area coding apparatus according to a second embodiment of the present invention.

The present embodiment is configured by adding an approximate character generator 114, a differentiator 116 and a prediction error encoder 118 to the character area encoding apparatus of FIG.

In this embodiment, the approximate character generation unit 114 determines the pixel value of the approximate character area from the font 107 and the start position 109 of the character portion, the character interval 110, the font type 111, the character size 112, and the character color 113. 11
5 is calculated in the same procedure as the character generation unit 201. This pixel value 115 is used as a predicted value of the image value of the character area 103 in the input image. The difference processing is performed by the differentiator 116, and a prediction error 117 is obtained. This prediction error 117
Is, for example, DPCM
And an encoded prediction error 119 is output.

Output from the encoding device as encoded data are the character code 105 and the start position 10 of the character part.
9, character spacing 110, font type 111, character size 112, character color 113, and encoding prediction error 11
9

FIG. 4 is a block diagram of a character area decoding apparatus corresponding to the character area encoding apparatus of FIG.

The character area decoding apparatus of this embodiment comprises a font database 202, an approximate character generation unit 205, a prediction error decoding unit 207, and an adder 209.

Various fonts 203 corresponding to the character codes 105 sent from the character area encoding apparatus are searched from the font database 202, and the fonts 203 are coded data (character part data) also sent from the character area encoding apparatus. Start position 109, character spacing 110, font type 1
11, the character size 112, and the character color 113) are input to the approximate character generation unit 208.

In the approximate character generation unit 205, the start value 109 of the character part, the character interval 110, the font type 111,
The character size 112 and the color 113 are decoded, and the pixel value 206 of the approximate character area is calculated in the same manner as the approximate character generation unit 114. The prediction error encoding unit 207 decodes the prediction error 119 sent from the character area encoding device. The pixel value 206 of the approximate character area and the decoded prediction error 208 are added by the adder 209, and the pixel 210 of the character area is output.

[0043]

As described above, according to the present invention, since knowledge about fonts is shared in transmission and reception, if characters in a character portion in an image are recognized, characters can be coded and reproduced. Therefore, the amount of information can be significantly reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram of a character area encoding device according to a first embodiment of the present invention.

FIG. 2 is a block diagram of a character region decoding device corresponding to the character region encoding device of FIG. 1;

FIG. 3 is a block diagram of a character area encoding device according to a second embodiment of the present invention.

FIG. 4 is a block diagram of a character region decoding device corresponding to the character region encoding device of FIG. 3;

FIG. 5 is a diagram illustrating an example of a conventional character area encoding method.

[Explanation of symbols]

 Reference Signs List 101 input image 102 character area extraction unit 103 character area 104 character recognition unit 105 character code 106 font database 107 various fonts corresponding to character code 105 108 character matching unit 109 start position of character part 110 character interval 111 font type 112 character Size 113 Character color 114 Approximate character generation unit 115 Pixel value of approximate character area 116 Difference unit 117 Prediction error 118 Prediction error encoding unit 119 Encoding prediction error 201 Character generation unit 202 Font database 203 Various fonts corresponding to character code 105 204 Pixel of character area 205 Approximate character generation unit 206 Pixel value of approximate character area 207 Prediction error decoding unit 208 Decoded prediction error 209 Adder 210 Pixel of character area

Claims (10)

[Claims] When encoding pixels of a character portion included in an image, the character portion is extracted as a character region, and after performing character recognition, character encoding is performed, and various characters corresponding to the character code are encoded. A character area encoding method for searching for a font closest to the pixel of the character portion of the font, and encoding parameters and parameters necessary for reproducing the character portion. 2. When encoding pixels of a character portion included in an image, the character portion is extracted as a character region, and after character recognition, character encoding is performed, and various characters corresponding to the character code are encoded. The font is searched for the font closest to the pixel of the character portion, and the parameters and parameters necessary for reproducing the character portion are encoded and output.The character font corresponding to the character code and the character font and both parameters are output. A character area encoding method for generating an approximate image of a part to be a predicted image for an input image, determining a prediction error, and encoding the prediction error. 3. The parameter of the font is a font type, a character size, and a character color, and parameters necessary for reproducing the character portion are a start position of the character portion and a character interval. 3. The character area encoding method according to 1 or 2. 4. A character area extracting means for extracting a character portion as a character area from an input image, a font database storing various fonts corresponding to the character codes, and recognizing characters from the character areas to convert the character codes. Output,
Character recognition means for searching the font database, inputting the font output from the character area and the font database, searching for a font closest to the pixel of the character area, and its parameters and reproducing the character part. A character region encoding device having character matching means for encoding and outputting necessary parameters. 5. A character area extracting means for extracting a character part as a character area from an input image, a font database storing various fonts corresponding to the character codes, and recognizing characters from the character areas to convert the character codes. Output,
Character recognition means for searching the font database, inputting the font output from the character area and the font database, searching for a font closest to the pixel of the character area, and its parameters and reproducing the character part. Character matching means for encoding and outputting necessary parameters; inputting both parameters output from the character matching means and a font output from the font database to generate an approximate character generating an approximate image of the character portion A character region encoding apparatus comprising: means, a difference unit that calculates a difference between the character region and the approximate image and outputs the difference as a prediction error, and a prediction error encoding unit that encodes the prediction error. 6. The font parameter is a font type, a character size, and a character color, and parameters necessary for reproducing the character portion are a start position of the character portion and a character interval. 6. The character area encoding device according to 4 or 5. 7. A character area decoding method corresponding to the character area encoding method according to claim 1, wherein a font corresponding to the character code is output, the parameters are decoded, and pixels of the character area are decoded. A character area decoding method to reproduce and output. 8. A character area decoding method corresponding to the character area encoding method according to claim 2, wherein a font corresponding to the character code is output, and an approximate image of the character part is obtained from the font and the two parameters. A character area decoding method for obtaining a pixel value, decoding the prediction error, adding the pixel value of the approximate image and the decoded prediction error, and reproducing and outputting a pixel of the character area. 9. A character area decoding apparatus corresponding to the character area encoding apparatus according to claim 4, wherein said font database stores various fonts and outputs a font corresponding to said character code. A character area decoding device comprising: a character generation unit that decodes the two parameters and reproduces the pixels of the character area from the decoded parameters and the font and outputs the pixels. 10. A character area decoding apparatus corresponding to the character area encoding apparatus according to claim 5, wherein: a font database storing various fonts and outputting a font corresponding to the character code; Approximate character generation means for decoding the approximate value of the image of the character portion from both the font and the decoded parameters; a prediction error decoding means for decoding the prediction error; and a pixel value of the approximate image and A character area decoding apparatus comprising an adding unit that adds decoded prediction errors, reproduces and outputs pixels of the character area.