JP4311663B2 - Document conversion apparatus, document conversion method, document conversion program, and recording medium - Google Patents

Description

  The present invention relates to a technique for realizing a high compression rate by forming an image corresponding to the type of an object on a plurality of planes and selecting one of the planes according to a gradation value given to a reference image.

  Conventionally, a technique is known in which an input image is separated into an edge portion and a non-edge portion, encoded with different encoding methods, and the output after encoding is switched according to the separation result (Japanese Patent Laid-Open No. 2003-134344). . In this image processing technique, only two switching between an edge portion and a non-edge portion is performed, and three or more are not supported.

  In addition, a binary image is generated from the input multivalued image, the binary image data is analyzed to extract a character portion, and pixels corresponding to the extracted character portion in the multivalued image are represented by pixel values of peripheral pixels. Replacement (Japanese Patent Application Laid-Open No. 2003-046746).

Replacing the multivalued image data corresponding to the black pixel data with the average color data does not reproduce the original image just by erasing the character area and creating an empty form.
JP 2003-134344 A JP 2003-046746 A

However, in the techniques of Patent Documents 1 and 2 , basically, the screen division is only divided into two types of characters and background. For example, encoding or the like is performed on three or more types of objects in the original image. I can't.

The purpose of the present invention, an image including hints or even background a plurality of types of objects, each document conversion device capable of realizing a high compression ratio by dividing into a plurality of planes, a document conversion method, a document conversion program And providing a recording medium in which the program is stored.

The present invention, from the original document has a plane image extracting means for extracting the two planes, the information for selecting any of the above two planes for each pixel as the value of two gradations, said code for encoding a reference image creation means for creating a plane and the reference image of two said reference image that contains a fill image of one tone of the two tone a predetermined object in the original on the document a document conversion apparatus that includes a means, the plane image extraction means, the resolution of the planes selected by the pixel of the fill image on the reference image to a lower resolution than the resolution of the original document extracting a region (hereinafter referred to as tiles) comprising a plurality of pixels to a predetermined resolution, in the tiles on the original image corresponding to the pixels on the plane Determining the variation in the pixel values of the fill image, when the variation is small, according to the average value of the pixel values of the fill image in the area, to determine the pixel values of the tiles on the plane image, the variation when is large, the size of the tile and more smaller, determines the variance of the pixel values of the fill image in a region smaller the size, when the variation is small, the average value of pixel values of the coating crush image Correspondingly, it is what determines the pixel values of the tiles on the plane image.

The plane image extraction means may extract the object on the original document by dividing it into the two planes, or extracting the object on the two planes and the reference image.

The encoding means may encode the two planes or the reference image by a method corresponding to the type of object on the original document.

The encoding means may encode at least one of the two planes by an encoding method different from the encoding method of the other planes.

It is preferable to provide integration means for integrating the reference image and the two planes encoded by the encoding means into a data set.

Further, the document conversion apparatus may include a processing unit that performs image processing or image correction.

Further, from the original document has a plane image extraction step of extracting a two planes, the information for selecting any of the above two planes for each pixel as the value of 2 gradations, said initial document coding step of coding the reference image creation step of creating, the plane and the reference image of two said reference image that contains a fill image of one tone of the two tone a predetermined object above a document conversion method having the bets, the plane image extracting step defines a resolution of the planes selected by the pixel of the fill image on the reference image in advance a lower resolution than the resolution of the original document was (hereinafter referred tile) multiple of pixels regions to a resolution to extract, the data on the original image corresponding to the pixels on the plane Determining the variation in the pixel values of the fill image in yl, when the variation is small, according to the average value of the pixel values of the fill image in the area, to determine the pixel values of the tiles on the plane image, when the variation is large, and smaller size of the tile, determines a variation in the pixel values of the fill image in a region smaller the size, when the variation is small, the average of the pixel values of the coating crush image depending on the value, it is what determines the pixel values of the tiles on the plane image.

In the plain image extraction step, the object on the original document may be extracted by dividing it into the two planes, or may be extracted separately on the two planes and the reference image.

In the encoding step, the two planes or the reference image may be encoded by a method corresponding to the type of object on the original document.

In the encoding step, at least one of the two planes may be encoded by an encoding method different from the encoding method of the other planes.

It is preferable to include an integration step of integrating the reference image encoded by the encoding step and the two planes into a data set.

The document conversion method may include a processing step for performing image processing or image correction.

Also provided is a document conversion program for causing a computer to function as each of the above means.

A computer-readable recording medium on which the document conversion program is recorded is also provided.

  According to the present invention, it is possible to reduce the file size of a document related to a color image without significantly reducing the image quality. Further, in a reproduced document, the visibility of characters, ruled lines, graphics composed of a single color, etc. can be maintained, and it can be applied to an image having gradation. In addition, since the document image is divided into a plurality of planes and a reference image, only necessary objects can be displayed or transmitted.

[First Embodiment]
FIGS. 1A and 1B are explanatory diagrams showing the document data structure of the first embodiment.

  As shown in FIG. 1B, the document data DD includes n (n = 2 in this embodiment) planes PLN (1) and PLN (2) extracted from the original document 100 shown in FIG. And the reference image REF are integrated.

  As shown in FIG. 1A, the original document 100 is composed of an 8-bit color image 101, a black filled character, and a gray filled character 102.

  In the present embodiment, the image displayed on the original document 100 is extracted by being divided into planes PLN (1) and PLN (2) according to the object described in the image as shown in FIG. Has been. The reference image REF has information for selecting one of the planes PLN (1) and PLN (2) for each pixel as values A1 and A2 of m gradations (m = 2 in this embodiment). Yes. For example, A1: “0” and A2: “1” can be set. In the present embodiment, the reference image REF includes a character image on the original document 100 that is a painted image of one gradation of m gradations. This filled image is a character in the present embodiment.

  In the present embodiment, the plane PLN (1) is, for example, an 8-bit color image and is encoded by a method suitable for compression of a photograph or the like.

  The plane PLN (2) is encoded by a method suitable for compression of a simple-shaped filled image, and the pixel value corresponding to the filled image (character) of the reference image REF is an object on the original document. It is determined by the pixel value of (character).

  The reference image REF is a character (an image image or an image represented by a text code) and is encoded by a method suitable for character compression.

  Since the plane PLN (1) is a color image, it is preferable to apply a compression method such as JPEG or JPEG2000. Further, the compression method such as JPEG or JPEG2000 can be applied to the plane PLN (2). Since the reference image REF is a character image, it is preferable to perform MMR compression.

  In this embodiment, as shown in FIGS. 1A and 1B, the resolution of the reference image REF is the same as the resolution of the original document 100, and the resolution of the plane PLN (2) is the resolution of the reference image REF. Is set lower.

  In order to reduce the low resolution of the plane PLN (2), it may be inconvenient to simply determine the pixel value of the image whose resolution has been reduced by thinning or averaging. For example, since the plane PLN (2) is an image for reproducing the color of a character, it is not preferable that pixel values other than characters are mixed into the plane PLN (2). That is, since the portion other than the characters is a pixel value that is appropriately determined in a single color such as black, it is not appropriate to reflect this pixel value in the image of the plane PLN (2) with reduced resolution.

  2A, 2B, and 2C are explanatory diagrams of the resolution reduction processing of the plane PLN (2). 2A shows the pixel values on the plane PLN (2), and FIG. 2B determines the pixel values on the plane PLN (2) from the pixel values of the reference image REF in FIG. FIG. 2C shows pixel values on the plane PLN (2). The average value calculation image GA is obtained by extracting only the pixel values of the character object of the original document 100, and the pixel values on the plane PLN (2) are 3 × 3 pixels of the average value calculation image GA. This is given as the average of the pixel values of the character objects included in the tile.

  In FIG. 2B, for the white pixel of the reference image REF, the pixel value of the plane PLN (1) is selected, and for the black pixel, the pixel value on the plane PLN (2) (the pixel value average of the character object). Value: dark gray in the figure) is selected.

  FIG. 3 is a block diagram illustrating the document conversion apparatus according to the first embodiment. Document data DD having the structure illustrated in FIG. 1B can be created by the document conversion apparatus.

  In FIG. 3, the document conversion apparatus 11 includes a plain image extraction unit 111, a reference image creation unit 112, a processing unit 113, an encoding unit 114, and an integration unit 115.

  The plane image extraction unit 111 can extract two planes PLN (1) and PLN (2) corresponding to the encoding method from the original document 100.

  In the present embodiment, the plane PLN (1) is a color image. In PLN (2), when the color of the object on the original document 100 is a color, the average value of the objects on PLN (2) is used. The reference image REF is a black and white image.

  In the present embodiment, a color photograph or the like is extracted to the plane PLN (1) and characters are extracted to the plane PLN (2). The distinction of objects (character extraction, background extraction) at the time of these extractions This technique is well known and will not be described.

  The reference image creation unit 112 can create a reference image REF.

  The processing means 113 can perform image processing or image correction (resolution conversion, filtering processing, lightness correction, color tone correction, noise removal). Normally, this processing increases the compression rate during encoding.

  For example, the plane PLN (2) only needs to know the character color on the restored document 100 '. Therefore, the resolution on the plane PLN (2) is set low. In addition, since the object on the reference image REF is a character, the visibility is not reduced much even if the resolution of the plane PLN (2) is reduced. A text code can be obtained by performing character recognition processing on the reference image REF. If this is associated with a series of data sets to be finally generated (usually, text data is included in the data set), the restored document 100 can be searched by text codes.

  Further, image processing for increasing the resolution may be performed by interpolation processing or the like. When the file size is large but the photo part is important, the image quality is improved. The reference image REF is not changed in the dot pitch in this embodiment, but can be changed.

  With respect to the reference image REF, noise can be removed by a method such as erasing the connected components of minute black pixels, thereby increasing the compression rate at the time of encoding.

  Note that the processing of each plane can be appropriately determined according to how the document is used. For example, if the photograph is considered important, the resolution of the plane PLN (1) should not be lowered. Conversely, if the photograph is not so important, the resolution of the plane PLN (1) should be lowered. However, the resolution of the plane PLN (2) may be increased. Since these are trade-offs between image quality and final information size, they are determined according to the application.

  The encoding unit 114 is a method according to the type of the object on the original document 100 (the method suitable for compression of the object, and the plane PLN (1), PLN (2)) and the reference image REF are different for each plane. ) Can be encoded.

  The integration unit 115 can integrate the reference image REF encoded by the encoding unit 113 and the n planes PLN (1) and PLN (2) into a data set. It should be noted that at the time of integration, information about what attributes each plane has (for example, bitmap images such as photographs, information on characters, backgrounds, etc.) is included in the data set, for example, as described later. At the time of document restoration, planes unnecessary for the operator can be excluded using the plane exclusion unit 213. As a result, only a specific plane can be seen in the image representation. In this case, unnecessary information becomes invisible, so that the document can be easily viewed. Further, when data communication is performed, planes that are considered unnecessary are appropriately excluded from the data set, so that the amount of data to be transmitted / received can be reduced (thus shortening the transmission / reception time).

  FIG. 4 is an explanatory diagram illustrating the document conversion method according to the first embodiment. This document conversion method can be implemented using the document conversion apparatus 11 of FIG.

  4, first, two planes PLN (1) and PLN (2) corresponding to the encoding method are extracted from the original document 100 (plane extraction step: S101). In this plane extraction step S101, the object on the original document 100 is extracted by dividing it into two planes PLN (1) and PLN (2).

Next, two gradation values A1 and A2 are assigned to each pixel, and a reference image for specifying one of the two planes PLN (1) and PLN (2) according to the two gradation values. A REF is created (reference image creation step: S102). ,
Further, image processing or image correction (resolution conversion, filtering processing, lightness correction, color tone correction, noise removal) is performed on the two planes PLN (1) and PLN (2) (processing step: S103). The resolution conversion described above is performed for the plane PLN (2).

  Thereafter, the two planes PLN (1), PLN (2) and the reference image REF are encoded (encoding step: S104). In the encoding step S104. The two planes PLN (1), PLN (2) and the reference image REF are encoded by a method according to the type of object (character in this embodiment) on the original document 100. In the present embodiment, the two planes are encoded by different encoding methods.

  Then, the reference image REF encoded in the encoding step S104 and the two planes PLN (1) and PLN (2) are integrated into a data set.

  Note that the pixels corresponding to the characters on the original document 100 in the plane PLN (1) can be replaced with representative representative colors. In addition, the pixel value of the unreferenced area of the plane PLN (2) can be replaced with a predetermined pixel value (for example, white, black, etc.).

  Thereby, the compression efficiency can be improved during encoding.

  FIG. 5 is an explanatory diagram illustrating the document restoration apparatus according to the first embodiment.

  In FIG. 5, the document restoration device 21 includes a document decomposition / decryption unit 211, a gradation value acquisition unit 212, a plane exclusion unit 213, a plane identification unit 214, and a document restoration unit 215. In the present embodiment, the document restoration device 21 can acquire the document data DD and restore the original document 100 from the document data.

  The document decomposition / decryption unit 211 can decrypt the document data DD and generate two plane planes PLN (1), PLN (2) and a reference image REF.

  The gradation value acquisition unit 212 scans the pixels of the reference image REF, and can acquire one of the two gradation values A1 and A2 given to the scanned pixels.

  The plane excluding unit 213 can exclude one or two of the two planes PLN (1) and PLN (2) from the processing target. The planes excluded by the plane excluding unit 213 are not included in the specific target by the plane identifying unit 214, so that only necessary objects can be displayed on the reproduced image 100 ′.

  The plane specifying unit 214 can specify the planes PLN (1) and PLN (2) corresponding to the two gradation values A1 and A2.

  The document restoration unit 215 extracts the value on the plane specified by the plane identification unit 214 corresponding to the pixel given to the pixel position of the reference image REF, restores the document, and restores the restored document 100 ′. Obtainable.

  When the color is defined on the plane PLN (2) and the gradation value of the pixel of the reference image REF is A2, the restored document 100 ′ reflects the color defined on the plane PLN (2). Is done.

  FIG. 6 is an explanatory diagram illustrating the document restoration method according to the first embodiment. This document restoration method can be implemented using the document restoration device 21 of FIG.

  In FIG. 6, first, document data DD is decomposed into two planes PLN (1), PLN (2) and reference image REF, and these are subjected to decoding processing (document decomposition / decoding step: S201).

  The pixels of the reference image REF are scanned to acquire the two gradation values A1 and A2 assigned to each pixel (gradation value acquisition step: S202).

  A plane to be a predetermined target is excluded (plane exclusion step: S203).

  A plane corresponding to the two gradation values A1 and A2 is specified (plane specifying step: S204).

The pixel value given to the pixel position on the plane specified in the plane specifying step S204 is extracted to generate the restored document 100 ′ (document restoring step: S205).
[Second Embodiment]
FIGS. 7A, 7 </ b> B, and 7 </ b> C are explanatory diagrams of the embodiment in which the resolution reduction processing of the plane PLN (2) is changed. FIG. 7A shows pixel values on the plane PLN (2), and FIG. 7B determines pixel values on the plane PLN (2) from the pixel values of the reference image REF in FIG. FIG. 7C shows pixel values on the plane PLN (2).

  The average value calculation image GA is obtained by extracting only the pixel values of the character object of the original document 100, and the pixel values on the plane PLN (2) are 8 × 8 pixels of the average value calculation image GA. This is given as the average of the pixel values of the character objects included in the tile T.

  In FIG. 7B, for the white pixel of the reference image REF, the pixel value of the plane PLN (1) is selected, and for the black pixel, the pixel value on the plane PLN (2) (the pixel value average of the character object) Value: dark gray in the figure) is selected.

  When the variation of pixels corresponding to characters in the tile T of 8 × 8 pixels on the average value calculation image GA is small, the values of the corresponding pixels on the plane PLN (2) are rounded to the same color. Further, when the pixel value variation in the tile T is large, the variation in the smaller size tile (for example, a tile of 2 × 2 pixels) is determined. Pixel values can be rounded to the same color.

  FIG. 8 is a flowchart showing processing for reducing the resolution of the plane PLN (2) described above.

  First, a tile T corresponding to 8 × 8 pixels is selected on the average value calculation image GA (S301). Next, a pixel corresponding to the character position in the tile T is extracted (S302), and it is determined whether there is a corresponding pixel (S303).

  In step S303, when there is no pixel corresponding to the character position in the tile T, a predetermined pixel value is assigned to the pixel of the plane PLN (2) (S304), and a pixel value is assigned to all the pixels of the plane PLN (2). It is determined whether or not (S310).

  If there is a pixel corresponding to the character position in the tile T in step S303, the average and variance of each pixel value are calculated (S305), and it is determined whether the variance is greater than a predetermined value (S306). . When the variance is larger than the predetermined value, the 8 × 8 pixel tile T on the average value calculation image GA is subdivided into 2 × 2 pixel tiles (S307), which correspond to the characters in the subdivided tiles. The average value of the pixel values is assigned to the pixels of the plane PLN (2) (S308), and it is determined whether or not the pixel values have been given for all the pixels of the plane PLN (2) (S310).

  If the variance is larger than the predetermined value in step S306, the average value of the pixel values corresponding to the characters in the tile T of 8 × 8 pixels is assigned to the PLN (2) pixels, and all of the planes PLN (2) are obtained. It is determined whether or not pixel value assignment has been completed for the pixel (S310).

  In S310, when the pixel value assignment is finished for all the pixels of the plane PLN (2), the resolution reduction processing is finished.

  In the present embodiment, the color variation of the original document 100 is determined, and if there is a variation, coloring is performed in finer units so that the colors are not mixed. In the present embodiment, since the coloring unit is selectively changed, an effect of increasing the compression efficiency can be expected without degrading the image quality.

In addition, as a method for obtaining the color variation, in addition to the dispersion method, a ratio of a pixel having an achromatic color and a pixel having a chromatic color may be used. In this case, it is determined that the color variation is large except when one of the chromatic color and the achromatic color is dominant. The achromatic color is determined when | R−G | <th and | GB | <th and a predetermined threshold th when the color of a pixel is (R, G, B). | B-R | <th
If the above relationship holds, a method of determining an achromatic color can be considered.

  In such a method using the ratio of achromatic color to chromatic color, if the chromatic colors are mixed, the colors will be mixed, but the multiplication required for the distributed calculation is unnecessary, so the cost for the calculation Has the advantage of being small.

It is explanatory drawing which shows the document data structure of 1st Embodiment, (A) is a figure which shows an original document, (B) is a figure which shows the document data by which each plane and the reference image were integrated. It is explanatory drawing of the resolution reduction process of plane PLN (2), (A) shows the value of the pixel on plane PLN (2), (B) is a plane from the pixel value of the reference image REF of (A). An average value calculation image GA for determining a pixel value on PLN (2) is shown, and (C) is a diagram showing pixel values on a plane PLN (2). It is a block diagram which shows the document conversion apparatus of 1st Embodiment. It is explanatory drawing which shows the document conversion method of 1st Embodiment. It is explanatory drawing which shows the document restoration apparatus of 1st Embodiment. It is explanatory drawing which shows the document restoration method of 1st Embodiment. FIG. 3 is an explanatory diagram of processing in which the resolution reduction processing of the plane PLN (2) in FIG. 2 is changed, in which (A) shows the values of pixels on the plane PLN (2), and (B) shows the processing of (A). An average value calculation image GA for determining the pixel value on the plane PLN (2) from the pixel value of the reference image REF is shown, and (C) is a diagram showing the pixel value on the plane PLN (2). . It is a flowchart which shows the process which lowers the resolution of plane PLN (2).

Explanation of symbols

11 Document Conversion Device 21 Document Restoration Device 100 Original Document
100 'restored document 101 color image 102 character 103 background 111 plain image extraction means 112 reference image creation means 113 correction means 114 encoding means 115 integration means 211 document decomposition / decoding means 212 gradation value acquisition means 213 plane exclusion means 214 plane Identification means 215 Document restoration means DD Document data A Tone value PLN (x) (x is a positive integer) Plane REF Reference image

Claims (14)

Plain image extraction means for extracting two planes from the original document;
Information for selecting one of the two planes for each pixel is provided as a two- gradation value, and a predetermined object on the original document is applied to one gradation of the two gradations. A reference image creating means for creating a reference image to be included as a crushing image;
A document conversion apparatus comprising: encoding means for encoding the two planes and the reference image;
The plane image extraction means includes a plurality of pixels so that the resolution of the plane selected by the pixels of the filled image on the reference image is lower than the resolution of the original document and has a predetermined resolution. Extract the area consisting of (hereinafter referred to as tiles)
Judge the variation of the pixel value of the filled image in the tile on the original image corresponding to the pixel on the plane, and when the variation is small, according to the average value of the pixel value of the painted image in the region, Determining pixel values of the tile on the plain image;
When the variation is large, reduce the size of the tile, determine the variation in the pixel value of the filled image in the region where the size is reduced,
When the variation is small, the document conversion apparatus determines a pixel value of a tile on the plain image according to an average value of pixel values of the filled image . 2. The plane image extraction unit extracts the object on the original document by dividing it into the two planes, or extracts the object on the two planes and the reference image. Document conversion device described in 1. The document conversion apparatus according to claim 1, wherein the encoding unit encodes the two planes or further the reference image by a method according to a type of an object on the original document. It said encoding means, at least one of the two planes, the document conversion apparatus of claim 1, characterized in that coded by different coding method and coding method other planes. The document conversion apparatus according to claim 1, further comprising an integration unit that integrates the reference image and the two planes encoded by the encoding unit into a data set.   The document conversion apparatus according to claim 1, further comprising a processing unit that performs image processing or image correction. A plain image extraction step of extracting two planes from the original document;
Information for selecting one of the two planes for each pixel is provided as a two- gradation value, and a predetermined object on the original document is painted with one gradation of the two gradations. A reference image creation step for creating a reference image to be included as a collapsed image;
A document conversion method comprising: an encoding step of encoding the two planes and the reference image,
The plane image extraction step includes a plurality of pixels so that a resolution of a plane selected by pixels of a painted image on the reference image is lower than a resolution of the original document and is a predetermined resolution. Extract the area consisting of (hereinafter referred to as tiles)
Judge the variation of the pixel value of the filled image in the tile on the original image corresponding to the pixel on the plane, and when the variation is small, according to the average value of the pixel value of the painted image in the region, Determining pixel values of the tile on the plain image;
When the variation is large, reduce the size of the tile, determine the variation in the pixel value of the filled image in the region where the size is reduced,
A document conversion method characterized in that, when the variation is small, a pixel value of a tile on the plain image is determined according to an average value of pixel values of the filled image. 8. The plane image extracting step is characterized in that an object on the original document is extracted by being divided into the two planes, or is extracted by being divided on the two planes and the reference image. Document conversion method described in 1. 8. The document conversion method according to claim 7, wherein the encoding step encodes the two planes or further the reference image by a method according to a type of an object on the original document. Said encoding step, said at least one of the two planes, the document conversion method according to claim 7, characterized in that coded by different coding method and coding method other planes. The document conversion method according to claim 7, further comprising an integration step of integrating the reference image encoded by the encoding step and the two planes into a data set.   8. The document conversion method according to claim 7, further comprising a processing step for performing image processing or image correction. A document conversion program for causing a computer to function as each means according to claim 1. Recorded computer read-usable medium the document conversion program according to claim 13.