JP2006279741A - Method, apparatus and program for image compression, and image reading apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method, apparatus and program for image compression and image reading apparatus with which a digital image is compressed with high compressibility so as not to damage readability of character. <P>SOLUTION: The image compression method includes: a setting stage of setting a processing target region on a digital image; a grouping stage of grouping pixels constituting the processing target region for each pixel with little color difference; a selecting stage of selecting a replacement target group based on feature amounts of the groups; a location information generating stage of indicating a location of each of the pixels belonging to the replacement target group; a replacing stage of replacing the color of each of pixels belonging to the replacement target group with the color of surrounding pixels; a digital image compressing stage of generating image compression data by compressing the digital image after the color of each of the pixels is replaced; and an output stage of outputting color data representing the representative color of the replacement target group, the location information and the image compression data in an associated manner. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image compression method, an image compression apparatus, an image compression program, and an image reading apparatus.

  There is known an image reading apparatus that reads a document to generate a digital image, compresses the generated digital image, and stores the compressed digital image on a removable memory card (see, for example, Patent Document 1). However, a memory card generally has a smaller storage capacity than a hard disk or the like provided in a personal computer. On the other hand, even if a digital image is compressed, the amount of data is still large. turn into.

  For example, when distributing a document to another person, a paper document may be read by an image reading device to generate a digital image, and the generated digital image may be attached to an e-mail and sent. However, as described above, even if a digital image is compressed, the amount of data is still large. Therefore, if it is attached to an e-mail and transmitted, it may take a long time to transmit or may not be transmitted due to a transmission error.

  Since a large amount of digital image data causes various problems as described above, a method capable of compressing the digital image smaller is desired. Since the compression rate of JPEG compression is variable, the amount of data can be reduced by compressing at a higher compression rate. However, JPEG compression is irreversible compression, and in irreversible compression, a larger amount of data is discarded as the compression rate is increased. Therefore, the image quality is lowered as the compression rate is increased. For example, in the case of a digital image generated by reading a document, if the compression ratio is increased, characters are blurred during reproduction and it becomes difficult to read. For this reason, a digital image generated by reading a document has a problem that the compression rate cannot be easily increased.

JP 2000-13577 A

  The present invention has been made in view of the above problems, and is an image compression method, an image compression apparatus, an image compression program, and an image reading apparatus for compressing a digital image at a high compression rate without impairing the readability of characters. The purpose is to provide.

(1) In order to achieve the above object, an image compression method according to the present invention includes a setting step of setting a processing target area on a digital image and a group of pixels constituting the processing target area for each pixel having a small color difference. A grouping step for selecting, a selection step for selecting a replacement target group based on a feature amount of each group, a position information generation step for generating position information indicating the position of each pixel belonging to the replacement target group, A replacement step of replacing the color of each pixel belonging to the replacement target group with a color of a surrounding pixel not belonging to the replacement target group; and the digital image after the replacement of the color of each pixel in the replacement step is compressed A digital image compression stage for generating compressed image data, color data representing a representative color of the replacement target group, the position information, and the compressed image data. Comprising an output step for attaching output, a.
When the pixels constituting the processing target region are grouped for each pixel having a small color difference, the group to which the pixel representing the character belongs often has a specific tendency in the feature amount. Therefore, if a feature amount is used, a group to which a pixel representing a character belongs can be specified with a high probability. When a group to which a pixel representing a character belongs is set as a replacement target group and the color of a pixel belonging to the replacement target group in the digital image is replaced with the color of a surrounding pixel, the character can be removed from the digital image. When characters are removed from the digital image, the digital image compression rate does not affect the readability of the characters, so the digital image can be compressed at a high compression rate. On the other hand, the position information indicating the position of the pixel representing the character does not need to have the density of each pixel, so a small amount of data is sufficient. Therefore, the data amount of the entire digital image can be reduced. At the time of reproduction, characters are reproduced without blurring by replacing the color of the pixel indicated by the position information in the digital image with the representative color represented by the color data. Therefore, according to the present invention, a digital image can be compressed at a high compression rate without impairing the readability of characters.

(2) The setting step includes a conversion step of dividing the digital image into a plurality of blocks and converting each of the blocks into frequency components, and region setting for setting the processing target region based on the high-frequency components of the blocks. Stages.
If an area that does not include characters is set as a processing target area, processing is wasted. When a region containing characters is converted to a frequency component, the high frequency component becomes a large value. Therefore, if the processing target region is set based on the high frequency component of each block, the region not including the character is wasted as being set as the processing target region. Can be reduced.

(3) The region setting step includes the step of setting the entire digital image as the processing target region, and the number of blocks having a high frequency component equal to or greater than a predetermined value for each row including the blocks in the processing target region. And dividing the processing target area by a row in which the number of blocks having a high frequency component equal to or greater than a predetermined value is less than a first threshold, and the high frequency component having a predetermined value for each column of the blocks in the processing target area And dividing the processing target area by a column in which the number of blocks having a high frequency component equal to or greater than a predetermined value is less than a second threshold, and the processing target area in a row A plurality of rectangular processing target regions may be set on the digital image by repeating the step of dividing the processing target region and the step of dividing the processing target region in a row. .
When a plurality of processing target areas are set, the feature amount has a specific tendency more prominently, so that a group to which pixels representing characters belong can be selected more reliably.

(4) In the grouping step, the pixels constituting the processing target region may be grouped by a k-means method.
When grouped by the k-means method, pixels can be grouped more appropriately.

(5) The feature amount may be the number of pixels belonging to the group, and the group having the second largest number of pixels may be selected as the replacement target group in the selection step.
In a digital image generated by reading a document, the number of pixels representing the background is generally the largest, and the number of pixels representing the character is the second largest. Therefore, by selecting a group having the second largest number of pixels, a group to which pixels representing characters belong can be selected.

(6) When a pixel group which is surrounded by pixels belonging to another group and is isolated from other pixels belonging to the same group is referred to as a small group, the feature amount is a value of the small group belonging to the group. In the selection step, the group having the largest number of small groups may be selected as the replacement target group.
In general, a character is composed of a number of straight lines and curves spaced apart from each other. That is, a character is composed of several small groups. Therefore, by selecting the group having the largest number of small groups, the group to which the pixel representing the character belongs can be selected.

(7) In the selection step, a plurality of the groups may be selected as the replacement target group.
When a character is represented by two or more colors, by selecting a plurality of groups, it is possible to reduce the remaining of the character in the digital image without selecting the group to which the pixel representing the character belongs as a replacement target group.

  (8) The position information includes a binary image obtained by binarizing pixels constituting the processing target area into pixels belonging to the replacement target group and pixels not belonging to the replacement target group, and the replacement target group It may be region information for specifying the position of the processing target region.

(9) In the output step, the color data, the position information, and the compressed image data may be stored and output in one PDF file.
Since a program for reproducing a PDF file is widely used, if a compressed digital image is stored in a PDF file, it can be reproduced by many apparatuses.

(10) In order to achieve the above object, an image compression apparatus according to the present invention groups setting means for setting a processing target area on a digital image and pixels constituting the processing target area for each pixel having a small color difference. Grouping means for selecting, selecting means for selecting a replacement target group based on the feature amount of each group, position information generating means for generating position information indicating the position of each pixel belonging to the replacement target group, A replacement unit that replaces the color of each pixel belonging to the replacement target group with a color of a surrounding pixel that does not belong to the replacement target group; and the digital image after the color of each pixel is replaced by the replacement unit. Digital image compression means for generating compressed image data, color data representing a representative color of the replacement target group, the position information, and the compressed image data And an output means for outputting put communication, the.
According to the present invention, a digital image can be compressed at a high compression rate without impairing the readability of characters.

(11) In order to achieve the above object, an image reading apparatus according to the present invention comprises reading means for reading a document and generating a digital image, and an image compression apparatus according to claim 10.
According to the present invention, a digital image can be compressed at a high compression rate without impairing the readability of characters.

(12) To achieve the above object, an image compression program according to the present invention groups setting means for setting a processing target area on a digital image and pixels constituting the processing target area for each pixel having a small color difference. Grouping means for selecting, selecting means for selecting a replacement target group based on the feature amount of each group, position information generating means for generating position information indicating the position of each pixel belonging to the replacement target group, A replacement unit that replaces the color of each pixel belonging to the replacement target group with a color of a surrounding pixel that does not belong to the replacement target group; and the digital image after the color of each pixel is replaced by the replacement unit. Digital image compression means for generating compressed image data, color data representing a representative color of the replacement target group, the position information, and the image compression data Causing a computer to function as output means for outputting in association with data.
According to the present invention, a digital image can be compressed at a high compression rate without impairing the readability of characters.

  The functions of the plurality of means provided in the present invention are realized by hardware resources whose functions are specified by the configuration itself, hardware resources whose functions are specified by a program, or a combination thereof. The functions of the plurality of means are not limited to those realized by hardware resources that are physically independent of each other.

Hereinafter, embodiments of the present invention will be described based on a plurality of examples.
(First Example)
FIG. 2 is a schematic diagram showing the internal structure of the image scanner 1 as the image reading apparatus according to the first embodiment of the present invention. The image scanner 1 is a so-called stand-alone image scanner 1 that does not require control by a personal computer or the like and can independently read a document, compresses a digital image generated by reading the document, and converts the compressed digital image into a PDF (Portable). (Document Format) format is stored in the removable memory.

The platen 10 is formed of a transparent plate such as a glass plate, and the document M is placed on a surface (document placement surface) 10a facing upward. The original M is a reflective original such as a printed document, a photograph, a book, or a transparent original such as a 35 mm strip film or a 35 mm mount film.
The document cover 11 is connected to the housing 12 so as to be swingable between a posture in which the document placement surface 10a of the platen 10 is opened by a hinge 36 and a posture in which the document placement surface 10a is covered.

  The surface light source unit 13 is provided on the document cover 11. The surface light source unit 13 includes a transparent document lamp 14, a diffusion plate, a reflection plate, and the like. The transparent document lamp 14 is a tube illumination device such as a white cold cathode fluorescent lamp (CCFL), and is arranged such that its longitudinal axis extends parallel to the central axis of the guide rod 16 of the carriage 15. The surface light source unit 13 is covered with a document mat 35 when a reflected document is read. When reading a transparent original, the original mat 35 is removed and the surface light source unit 13 is exposed.

  The carriage 15 is accommodated in the housing 12. The carriage 15 is slidably fitted to a guide rod 16 parallel to the document placement surface 10a. The carriage 15 includes a reflective document lamp 17, a mirror 18, a lens 19, and an image sensor 20. The carriage 15 is pulled by a belt hung on a pulley driven by a carriage transport motor 33 (see FIG. 3), and reciprocates in the X direction parallel to the document placement surface 10a. The reflective original lamp 17 is a tube illumination device such as a CCFL. The mirror 18 and the lens 19 cause the reflected light of the reflective original illuminated by the reflective original lamp 17 and the transmitted light of the transparent original illuminated by the transparent original lamp 14 to enter the light receiving surface of the image sensor 20. When the reflective document lamp 17, the mirror 18, the lens 19, and the image sensor 20 are conveyed in the X direction by the carriage 15, the main scanning line on the document moves.

  The image sensor 20 is a so-called linear image sensor in which a large number of cells (not shown) composed of photodiodes are linearly arranged. The image sensor 20 is mounted on the carriage 15 so that a large number of cells are arranged in parallel with the central axis of the tubular reflecting original lamp 17. Charges corresponding to the amount of received light are accumulated in each cell of the image sensor 20 by photoelectric conversion. The electric charge accumulated in each pixel is transferred to the output unit of the image sensor 20 by a CCD (Charge Coupled Device). This charge transfer may be performed by a CMOS (Complementary Metal Oxide Semiconductor). The image signal output from the image sensor 20 is output to the AFE unit 26 (see FIG. 3).

FIG. 3 is a block diagram illustrating a main configuration of hardware of the image scanner 1.
The CPU 21 controls various units of the image scanner 1 by executing various programs stored in the ROM 22. The CPU 21 also functions as an image compression device by executing an image compression program stored in the ROM 22. The ROM 22 is a non-volatile memory that stores various programs, various data, and the like. These various programs and various data are stored in the ROM 22 by downloading from a predetermined server via a network, reading from a computer-readable storage medium such as the removable memory 23, and the like. The RAM 24 is a memory that temporarily stores various programs and various data. The RAM controller 25 controls data transfer between the CPU 21, the AFE unit 26, the image processing unit 27, the external interface controller 28, and the like and the RAM 24.

The transparent document lamp controller 29 is composed of an inverter circuit, a control circuit, and the like, and controls turning on and off of the transparent document lamp 14.
The reflection document lamp controller 30 includes an inverter circuit, a control circuit, and the like, and controls turning on and off of the reflection document lamp 17.
The image sensor controller 31 is a drive circuit that outputs drive pulses necessary for driving the image sensor 20 to the image sensor 20, and includes, for example, a synchronization signal generator and a drive timing generator.

The carriage controller 32 includes a carriage transport motor drive circuit, a control circuit, and the like. The carriage controller 32 receives a carriage conveyance control signal transmitted from the CPU 21 and controls the rotation direction, rotation speed, and the like of the carriage conveyance motor 33.
The analog front end (AFE) unit 26 stores the digitized scan image in the RAM 24 via the RAM controller 25 by performing level adjustment processing by adjusting the gain of the electric signal, quantization processing, and the like.

The image processing unit 27 performs signal processing for performing image processing such as gamma correction, shading correction, and image forming processing for forming a digital image represented in the YCbCr color space on the scan image stored in the RAM 24. DSP (Digital Signal Processor) performed in cooperation with
The removable memory controller 34 is an output mechanism that transfers data stored in the RAM 24 to a removable memory 23 as a nonvolatile storage medium connected to the card connector.
The external interface controller 28 connects the image scanner 1 and an external system such as a PC (Personal Computer) so that they can communicate with each other.

Next, an image compression program will be described.
FIG. 4 is a block diagram showing a logical configuration of the image compression program.
The setting unit 41 is a module that causes the CPU 21 to function as setting means.
The grouping unit 42 is a module that causes the CPU 21 to function as a grouping unit.
The selection unit 43 is a module that causes the CPU 21 to function as selection means.
The position information generation unit 44 is a module that causes the CPU 21 to function as position information generation means.

The replacement unit 45 is a module that causes the CPU 21 to function as replacement means.
The image compression unit 46 is a module that causes the CPU 21 to function as digital image compression means.

The output unit 47 is a module that causes the CPU 21 to function as output means.

Next, processing of the CPU 21 that executes the image compression program will be described.
5 and 6 are flowcharts showing the flow of processing of the CPU 21 that executes the image compression program, and FIG. 1 is a schematic diagram showing the outline of the processing. This process is started when the image scanner 1 reads a document and generates a digital image.
In S105, the CPU 21 executes the setting unit 41, divides the digital image into blocks of 8 × 8 pixels, and performs discrete cosine transform (DCT transform) on the Y component (luminance component) for each block, so that each block has a frequency. Convert to component. The block is converted into an 8 × 8 frequency coefficient (DCT coefficient) by DCT conversion. In the DCT coefficient, the upper left is a direct current component, and the other is an alternating current component. The AC component represents an AC component (high frequency component) having a higher frequency from the upper left toward the lower right.

FIG. 7 is a schematic diagram illustrating an example of a digital image generated by a document scanner reading a document. FIG. 8 is a schematic diagram showing an image obtained by DCT transforming the digital image shown in FIG. It is the schematic diagram shown by the square.
In S110 illustrated in FIG. 5, the CPU 21 executes the setting unit 41 and initially sets the entire DCT-converted digital image as a processing target area.

In S115, the CPU 21 executes the setting unit 41, and executes area division processing on the set processing target area. In the area division process, a plurality of rectangular process target areas are set on the digital image by repeating the horizontal division process and the vertical division process.
First, the horizontal division process will be described. In the horizontal division process, the line with less high frequency components than the first threshold is defined as a “separation line”, and the processing target area is horizontally aligned with the processing target area above and below the “separation line”. To divide. The “separation line” itself is excluded from the processing target area. For example, in the example shown in FIG. 9, if the first threshold is 3, the third row of the processing target region 51 is identified as a “separation row” because the number of blocks containing high-frequency components is 2. As a result, the processing target area 51 is divided into a processing target area 52 composed of the first to second lines and a processing target area 53 composed of the fourth to sixth lines. Here, the first threshold is a value set according to the width of the processing target area. The first threshold value may be, for example, a ratio of the number of blocks including high-frequency components in the total number of blocks constituting the row, and may be a “separated row” if the ratio is less than the first threshold value. In that case, the ratio may be constant regardless of the width of the processing target area. The first threshold value is a design item that can be selected as appropriate, and may be adjusted so that the processing target region is appropriately set. Details of the horizontal division process will be described below.

FIG. 10 is a flowchart showing the flow of horizontal division processing.
In S205, the CPU 21 determines whether or not the width of the processing target area is equal to or larger than a predetermined width. If it is equal to or larger than the predetermined width, the process proceeds to S210. When it is less than the predetermined width, the process proceeds to S260.
In S210, the CPU 21 sets the weighting coefficient and the first threshold described above based on the width of the processing target area. In the image scanner 1, a block that should not contain a high-frequency component may contain a high-frequency component due to the influence of dust at the time of reading, so that a line that should be a delimiter line may not be specified as a delimiter line. . For this reason, the CPU 21 determines whether or not the block (dust block) includes a high-frequency component due to the influence of dust in S235 described later. The weighting coefficient is used at that time. Hereinafter, an example of the weighting coefficient will be described.

  FIG. 11 is a schematic diagram illustrating an example of a weighting coefficient. In FIG. 11A, 1 is set in the central block of interest and the surrounding 8 blocks. The CPU 21 sets the block value of the block including the high frequency component to 1, the block value of the block not including the high frequency component to 0, obtains the product of the block value and the weighting coefficient for each block, and sums them (block determination value) ) Is less than a predetermined value (dust determination value), the target block is determined to be a dust block. When there is a block containing only one high-frequency component where there is no block containing high-frequency components in the surrounding area, there is a high possibility that the block of interest is a dust block. According to the weighting coefficient shown in FIG. 11A, even if the block of interest includes a high frequency component, it is regarded as a dust block if there is no block including a high frequency component around it. In FIG. 11B, 1 is set for the block of interest and the four blocks on the top, bottom, left, and right. In FIG. 11C, 1 is set only for the block of interest. In the weighting coefficient shown in FIG. 11C, surrounding blocks are not taken into consideration in determining whether the block of interest is a dust block. That is, in the case of the weighting coefficient shown in FIG. 11C, it is not necessary to calculate a block determination value. If the target block includes a high-frequency component, it is determined as a block including a high-frequency component, and if it does not include a high-frequency component. It is determined that the block does not contain a high frequency component. For example, if the area of interest is wide, there will be no significant influence on the separation line even if there is one or two garbage, but if the area of interest is narrow, one garbage will greatly affect the identification of the separation line. . Therefore, for example, when the width of the attention area is wide, the process of calculating the block determination value can be omitted by using the weighting coefficient of FIG. 11C, and when the width of the attention area is narrow, the weighting coefficient of FIG. By using it, it is possible to specify the separator line more accurately. In the middle, the weighting coefficient shown in FIG. 11C may be used.

In S215 shown in FIG. 10, the CPU 21 sets the first row of the processing target area as the attention row. A row here is not a row in units of pixels but a row in units of blocks, and is a row in which blocks are arranged in a straight line.
In S220, the CPU 21 sets 0 to a counter N that counts the number of blocks containing high frequency components.
In S225, the CPU 21 selects one block of the target row and sets it as the target block. In this embodiment, the blocks are selected in order from the left end.
In S230, the CPU 21 determines whether or not the block of interest includes a high frequency component. Specifically, for example, if a value obtained by weighting the values of the six alternating current components 55 shaded in FIG. 12 with a predetermined weighting coefficient is equal to or greater than a predetermined value (high frequency determination value), the block is a high frequency component. It is determined that the block includes. The six AC components 55 correspond to the high frequency components described in the claims. Note that it is a design matter that can be appropriately selected to define the AC component up to which range is defined as a high frequency component based on the AC component in the lower right. For example, the range in which the processing target region is determined to be most appropriately set It may be obtained by experiment. For example, the high frequency component may be a total of three components, the lower right AC component and the AC component on the upper right, and the left AC component, or only the lower right AC component as the high frequency component. Also good. In any case, at least the lower right AC component is a high frequency component. Further, the weighting coefficient, the high-frequency determination value, and the like are design items that can be appropriately selected, and may be adjusted so that the processing target region is appropriately set. If the high frequency component is included, the process proceeds to S235, and if not, the process proceeds to S245.

  In S235, the CPU 21 determines whether or not the target block is a dust block. Specifically, the block determination value is calculated based on the weighting coefficient set in S210, and it is determined whether or not the block determination value is greater than or equal to the dust determination value. If it is greater than or equal to the dust determination value, it is determined that the block is not a dust block, and the process proceeds to S240. If it is less than the dust determination value, it is determined as a dust block and the process proceeds to S245.

In S240, the CPU 21 adds 1 to the counter N.
In S245, the CPU 21 determines whether there is a block on the right side of the block of interest. If it exists, the process returns to S225, and if it does not exist, the process proceeds to S250.
In S250, the CPU 21 determines whether or not the target line is a separator line. Specifically, it is determined whether or not the number (counter N) of blocks including high frequency components included in the target row is less than the first threshold value. If it is less than the first threshold, it is determined that the target line is a delimiter line, and the process proceeds to S255. If it is equal to or greater than the threshold, the process proceeds to S265.

In S255, the CPU 21 sets up to one line above the target line as one processing target area. However, if there is no line above or if the line above is a separator line, the processing target area is not set.
In S260, the CPU 21 executes vertical division processing on the processing target area set in S255.
In S265, the CPU 21 determines whether or not the target line is the last line. If it is not the last line, the process returns to S215. If it is the last line, the horizontal division process is terminated.

Next, the vertical division process will be described.
FIG. 13 is a flowchart showing the flow of the vertical division process. The vertical division process is substantially the same as the horizontal division process except that the rows are columns, the horizontals are vertical, the widths are high, and the first threshold is the second threshold. In the vertical division process, the horizontal division process is recursively called in S360. The horizontal division process recursively called from the vertical division process recursively calls the vertical division process in S260. Thus, the CPU 21 recursively calls the horizontal division process and the vertical division process until the width of the processing target area is less than the predetermined width or less than the predetermined height, thereby processing a plurality of rectangles on the digital image. Set the target area.

  FIG. 14 is a schematic diagram showing the result of the above-described area division processing. As a result of the region division processing described above, the digital image shown in FIG. 8 is divided into a plurality of processing target regions 56 as shown in FIG. In FIG. 14, each of a plurality of rectangular areas with parallel diagonal lines corresponds to one processing target area. According to the region dividing process described above, a region composed of blocks that do not include a high-frequency component is a rectangular region that is not attached with parallel diagonal lines. Since the area that does not include characters should not have a block that includes high-frequency components, it is a rectangular area that is not marked with parallel diagonal lines. There is no problem even if the area containing no characters is compressed as it is at a high compression rate. Nevertheless, if an area that does not include characters is set as the processing target area, processing is wasted. According to the area dividing process described above, an area that does not include characters is excluded from the process target area, so that useless processes can be reduced. Although the processing target area of the digital image after DCT conversion is shown here, the processing target area is set on the original digital image that is not actually DCT converted. In the following description, the “processing target area” refers to a processing target area set on the original digital image that has not been DCT transformed.

In S120 illustrated in FIG. 5, the CPU 21 selects one processing target area.
In S125, the CPU 21 executes the grouping unit 42 and executes a grouping process for grouping the pixels constituting the selected processing target area for each pixel having a small color difference. Details of the grouping process will be described below.
FIG. 15 is a flowchart showing the flow of grouping processing. Specifically, the grouping unit 42 performs grouping by, for example, a known k-means method. The k-means method is a method for appropriately grouping a plurality of elements.

In S405, the CPU 21 converts the color space of the processing target area into a color space independent of the model. Specifically, for example, conversion to the Lab color space is performed. In addition, you may convert into XYZ color space.
In S410, the CPU 21 divides the Lab color space into equal N ³ spaces. Here, a case where N = 3 will be described as an example. When N = 3, the Lab color space is divided into 27 equal spaces as shown in FIG. FIG. 16 shows a case where each component is represented by a gradation value of −128 to 127.

In S415, the CPU 21 sets the color specified by the positions of the vertices of 27 spaces as a temporary representative color. When N = 3, the number of vertices is (N + 1) ³ (= 64). Since the distances between adjacent vertices are all constant, the tentative representative color can be set in the entire Lab color space without bias by setting the color specified at the vertex position as the tentative representative color.
In S420, the CPU 21 obtains a color difference from the 64 temporary representative colors for each pixel in the processing target area, and assigns each pixel to the same group as the representative color having the smallest color difference. The color difference is obtained by the following formula 1, where ΔE is the color difference, (L ₁ , a ₁ , b ₁ ) is the color of a certain pixel, and (L ₂ , a ₂ , b ₂ ) is a representative color.

As a result, 64 temporary groups are formed.
In S425, the CPU 21 obtains the center of gravity for each group, and uses the obtained center of gravity as a new temporary representative color of each group. Here, the center of gravity means a color represented by (average value of L components of pixels belonging to a group, average value of a components of pixels belonging to a group, average value of b components of pixels belonging to a group).

In S430, the CPU 21 obtains the color difference between the old and new representative colors for each group.
In S435, the CPU 21 determines for each group whether the color difference between the old and new representative colors is equal to or greater than a predetermined threshold. If the color difference between the new and old representative colors is greater than or equal to a predetermined threshold in one or more groups, the process returns to S420 and is repeated. When the above-described processing is repeated, the center of gravity eventually converges, and the color difference between the new and old representative colors is less than a predetermined threshold value in any group. If the color difference is less than the predetermined threshold value, the process proceeds to S440 with the temporary representative color as the official representative color.

In S440, the CPU 21 determines whether or not the processing target area is a text area representing a character string. In the case of a text area, as shown in FIG. 17A, the pixels should be concentrated in two groups, a group to which pixels representing the background belong and a group to which pixels representing the character belong. Therefore, if the pixels are concentrated in any two groups, it can be said that there is a high possibility that the processing target area is a text area. Conversely, if the pixels are scattered in many groups as shown in FIG. 17B, it can be said that there is a high possibility that the processing target area is an image area representing an illustration or a photograph. Specifically, for example, if 80% or more of the whole is concentrated in two groups, the CPU 21 determines that it is concentrated in two groups and proceeds to S445. If it is less than 80%, the CPU 21 ends the grouping process. To do. Note that whether or not 80% is used as a criterion is a design item that can be selected as appropriate. In addition, here, the case where it is determined that it is a text area if concentrated in any two groups has been described as an example. However, since characters are often printed in two or more colors, for example 3-4 If it is concentrated in a group, it may be determined that it is a text area.
In S445, the CPU 21 adds a flag indicating that the processing target area is a text area, and ends the grouping process.

In S130 shown in FIG. 5, the CPU 21 determines whether or not the processing target area is a text area based on the flag. If it is a text area, the process proceeds to S135, and if it is not a text area, the process proceeds to S155.
In S135, the CPU 21 executes the selection unit 43 and executes a selection process for selecting a replacement target group in the processing target area based on the feature amount of each group in the processing target area. Here, the number of pixels belonging to each group will be described as an example of the feature amount. The selection unit 43 selects a group having the second largest number of pixels as a replacement target group. The reason for selecting the group having the second largest number of pixels is that the pixels belonging to the second largest group tend to be pixels representing the characters themselves. For example, the most common color in the processing target area indicated by the broken line 57 in FIG. 7 is gray indicating the background of the character. The next most common color is white, which represents the character itself. In this way, in a general document, the background has the largest number of pixels, and the text has the second largest number. However, this does not always hold, and if the background or character is printed in more than one color, the number of pixels representing the character itself may be first or third. In this way, since the ranking differs depending on the document, it cannot be said unequivocally, but on average, the probability of being the second is higher than the probability of being any other ranking. As described above, since characters may be printed in two or more colors, in addition to the second most common group, for example, the third most common group and the fourth most common group are selected as the replacement target group. May be. When a character is represented by two or more colors, by selecting a plurality of groups, it is possible to reduce the remaining of the character in the digital image without selecting the group to which the pixel representing the character belongs as a replacement target group.

  In the first embodiment, the digital image is divided into a plurality of processing target areas. However, if the digital image is divided into a plurality of processing target areas, there is an advantage that the feature amount has a specific tendency more prominently. For example, there are black characters and red characters, and the areas containing the black characters and the areas containing the red characters are set as different processing target areas. It is assumed that the processing target area is the second most common group. However, if those areas are set as one processing target area, even if one becomes the second largest group, the other will become the third largest group, only the second largest group. Is selected as a replacement target group, pixels belonging to the other group remain on the digital image. If the processing target area is divided into a plurality of groups, the group to which the pixel represented by the character belongs in any processing target area is the second most frequent, and the group to which the pixel representing the character belongs is the second most significant, Characters can be removed more reliably.

  In S140, the CPU 21 executes the position information generation unit 44 to generate position information indicating the position of each pixel belonging to the replacement target group. Specifically, the CPU 21 generates, for example, a bitmap in which pixels constituting the processing target area are binarized into pixels belonging to the replacement target group and pixels not belonging to the replacement target group. Hereinafter, the bitmap will be described.

  FIG. 18A is a schematic diagram of a certain processing target area, and FIG. 18B is a schematic diagram of a bitmap. In FIG. 18A, only the pixels belonging to the replacement target group are indicated by black frames. The processing target area is composed of n × m pixels, whereas the bitmap is composed of n × m bits. Each bit of the bitmap corresponds to each pixel in the processing target area, and a bit corresponding to a pixel belonging to the replacement target group is set to 1, and a bit corresponding to a pixel not belonging to the replacement target group is set to 0. In FIG. 18B, only the bits for which 1 is set are indicated by a black frame. In addition to the group with the second largest number of pixels, for example, when the third largest group and the fourth largest group are selected as the replacement target group, a bitmap is generated for each group. Note that the position information may be information indicating the coordinates of each pixel belonging to the replacement target group instead of the bitmap.

In S145 shown in FIG. 5, the CPU 21 compresses the generated bitmap by a compression method suitable for 1 bit such as G4, JBIG, Run Length, and generates bitmap compressed data.
In S150, the CPU 21 executes the replacement unit 45, and executes a replacement process for replacing the color of the pixel belonging to the replacement target group with the color of the surrounding pixels not belonging to the replacement target group in the digital image. Details of the replacement process will be described below.

  FIG. 19 is a schematic diagram for explaining the replacement process. FIG. 19 shows a digital image after the replacement process is executed for all the processing target areas. For example, it is assumed that the pixels constituting the character “su” in the processing target area indicated by the broken line 57 in FIG. In this case, since the character “su” is surrounded by gray pixels, the replacement unit 45 replaces the color of the pixel representing the character “su” with gray. The same applies to other characters. As a result, characters are removed from the processing target area as shown in FIG. However, the above-described example is an example in which the color of the adjacent pixel is only one color, and when the background is multicolored, the periphery is surrounded by a plurality of colors. In this case, only one of the surrounding colors may be replaced, or the nearest neighboring pixels may be specified and replaced for each pixel.

In S155 shown in FIG. 5, the CPU 21 determines whether or not the next processing target area exists. If it exists, the process returns to S120, and if it does not exist, the process proceeds to S160.
In S160 shown in FIG. 6, the CPU 21 executes the image compression unit 46, compresses the digital image after the color replacement with a high compression rate, and generates compressed image data. Specifically, for example, the CPU 21 JPEG compresses the digital image using a parameter for compressing at a high compression rate. As described above, since characters are removed from the digital image, there is no problem that the characters are difficult to read even when compressed at a high compression rate.

In S165, the CPU 21 executes the output unit 47, and color data representing the representative color of the replacement target group for each processing target area, bitmap compressed data generated by compressing the bitmap, and replacement as shown in FIG. The area information for specifying the position of the process target area to which the target group belongs is stored in a PDF file (PDF file), and the compressed image data generated by compressing the digital image is stored in the PDF file. The memory controller 34 is controlled and recorded in the removable memory 23. Since a program for reproducing a PDF file is widely used, if a compressed digital image is stored in a PDF file, it can be reproduced by many apparatuses. Here, the region information is specifically coordinates indicating a processing target region, for example. For a processing target area in which a plurality of groups are selected as replacement target groups, color data and bitmap compressed data are stored in association with each other for each replacement target group. In this embodiment, the representative color for the grouping process is the representative color for storing in the PDF file, but the representative color for storing in the PDF file is not necessarily the representative color for the grouping process. Also good. For example, when storing in a PDF file, the color of a pixel randomly selected from a group may be used as the representative color of the group.
At the time of reproduction, by associating the bitmap with the region on the digital image based on the region information, by replacing the color of the pixel specified by the bitmap with the color specified by the color data, The characters are played without bleeding.

  According to the image scanner 1 according to the first embodiment of the present invention described above, the replacement target group is selected based on the feature amount of each group. When the pixels constituting the processing target area are grouped for each pixel having a small color difference, the group to which the pixel representing the character belongs has the second largest number of pixels. Therefore, if a feature amount is used, a group to which a pixel representing a character belongs can be specified with a high probability. If a group to which a pixel representing a character belongs is set as a replacement target group and the color of a pixel belonging to the replacement target group in the digital image is replaced with the color of a surrounding pixel, the character can be removed from the digital image. When characters are removed from the digital image, the digital image compression rate does not affect the readability of the characters, so the digital image can be compressed at a high compression rate. On the other hand, since the bitmap need not have the density of each pixel, a small amount of data is sufficient. Therefore, the data amount of the PDF file can be reduced. Therefore, according to the image scanner 1, a digital image can be compressed at a high compression rate without impairing the readability of characters.

(Second embodiment)
The feature amount of the second embodiment is the number of small groups belonging to the group. Hereinafter, the processing target area indicated by the broken line 57 in FIG. 7 will be described as an example for the small group.
A small group refers to a pixel group surrounded by pixels belonging to another group and isolated from other pixels belonging to the same group. For example, in the processing target area indicated by the broken line 57, all the pixels representing the background are assigned to the same group, and all the pixels representing the characters are assigned to the same group different from the pixels representing the background. In the processing target area indicated by the broken line 57, there is no isolated pixel among the pixels representing the background, and therefore the whole corresponds to one small group.

  On the other hand, for example, the letter “di” is divided into five small groups in total. Therefore, many small groups belong to a group to which pixels representing characters belong, compared to a group to which pixels representing a background belong. In this way, in the processing target region with many katakana and hiragana, it can be said that the number of small groups of the group to which the pixel representing the character belongs is larger than the number of small groups of the group to which the pixel representing other than the character belongs. Therefore, by focusing on the number of small groups, the group to which the pixel representing the character belongs can be specified with high probability. However, this does not always hold. For example, if there are many kanji characters or a checkered background, the number of small groups may be smaller in the group to which the pixel representing the character belongs. In this way, the number of small groups differs depending on the sentence, so it can not be said unconditionally, but in the case of sentences that contain a lot of katakana and hiragana, the probability of being the first is higher than the probability of any other rank .

The second embodiment is substantially the same as the first embodiment in other points.
In the above-described embodiments, the image scanner has been described as an example of the image compression apparatus. However, the present invention may be applied to a personal computer. Further, the present invention may be applied to a digital camera, a camera-equipped mobile phone, and the like.

1 is a schematic diagram showing an outline of an image compression method according to a first embodiment of the present invention. 1 is a schematic diagram of an image reading apparatus according to a first embodiment of the present invention. 1 is a block diagram of an image reading apparatus according to a first embodiment of the present invention. 1 is a block diagram of an image compression program according to a first embodiment of the present invention. The flowchart which concerns on the 1st Example of this invention. The flowchart which concerns on the 1st Example of this invention. 1 is a schematic diagram of a digital image according to a first embodiment of the present invention. FIG. 3 is a schematic diagram obtained by converting a digital image according to the first embodiment of the present invention. The schematic diagram of the area | region division which concerns on 1st Example of this invention. The flowchart which concerns on the 1st Example of this invention. (A), (B) and (C) are schematic diagrams of weighting factors according to the first embodiment of the present invention. The schematic diagram of the frequency coefficient which concerns on the 1st Example of this invention. The flowchart which concerns on the 1st Example of this invention. The schematic diagram of the result of the area | region division which concerns on 1st Example of this invention. The flowchart which concerns on the 1st Example of this invention. FIG. 3 is a schematic diagram in which the color space according to the first embodiment of the present invention is divided equally. The graph of the feature-value which concerns on the 1st Example of this invention. (A) is a schematic diagram of the process target area | region which concerns on 1st Example of this invention, (B) is a schematic diagram of a binary image. The schematic diagram for demonstrating the substitution which concerns on the 1st Example of this invention.

Explanation of symbols

1 Image scanner (image reading apparatus), 21 CPU (image compression apparatus, setting means, grouping means, selection means, position information generation means, replacement means, digital image compression means, output means)

Claims (12)

A setting stage for setting the processing target area on the digital image;
A grouping step of grouping pixels constituting the processing target area for each pixel having a small color difference;
A selection step of selecting a replacement target group based on the feature amount of each of the groups;
A position information generation step of generating position information indicating the position of each pixel belonging to the replacement target group;
A replacement step of replacing the color of each pixel belonging to the replacement target group with the color of surrounding pixels not belonging to the replacement target group;
A digital image compression step of compressing the digital image after the color of each pixel is replaced in the replacement step to generate compressed image data;
An output step of associating and outputting color data representing a representative color of the replacement target group, the position information, and the compressed image data;
An image compression method comprising: The setting step includes:
Converting the digital image into a plurality of blocks, and converting each of the blocks into frequency components;
A region setting step for setting the processing target region based on the high-frequency component of each block;
The image compression method according to claim 1, further comprising: The region setting step includes:
Setting the entire digital image as the region to be processed;
The number of blocks having a high frequency component equal to or higher than a predetermined value is counted for each row including the blocks in the processing target area, and the number of blocks having a high frequency component equal to or higher than a predetermined value is less than a first threshold. Dividing the processing target area;
The number of blocks having a high frequency component equal to or greater than a predetermined value is counted for each column including the blocks for the processing target region, and the number of blocks having a high frequency component equal to or greater than a predetermined value is a column less than a second threshold value. Dividing the region to be processed,
A plurality of rectangular processing target areas are set on the digital image by repeating the step of dividing the processing target area by rows and the step of dividing the processing target area by columns. Item 3. The image compression method according to Item 2.   The image compression method according to claim 1, 2, or 3, wherein in the grouping step, pixels constituting the processing target region are grouped by a k-means method. The feature amount is the number of pixels belonging to the group,
5. The image compression method according to claim 1, wherein in the selection step, the group having the second largest number of pixels is selected as the replacement target group. When a pixel group surrounded by pixels belonging to another group and isolated from other pixels belonging to the same group is called a small group, the feature amount is the number of the small groups belonging to the group. ,
5. The image compression method according to claim 1, wherein, in the selection step, the group having the largest number of the small groups is selected as the replacement target group.   The image compression method according to claim 1, wherein in the selection step, a plurality of the groups are selected as the replacement target group.   The position information includes a binary image obtained by binarizing pixels constituting the processing target area into pixels belonging to the replacement target group and pixels not belonging to the replacement target group, and the processing target to which the replacement target group belongs. The image compression method according to claim 1, wherein the image compression method is region information for specifying a region position.   9. The image compression method according to claim 1, wherein in the output step, the color data, the position information, and the compressed image data are stored in a single PDF file and output. 10. . Setting means for setting a processing target area on the digital image;
Grouping means for grouping pixels constituting the processing target area for each pixel having a small color difference;
Selecting means for selecting a replacement target group based on the feature amount of each of the groups;
Position information generating means for generating position information indicating the position of each pixel belonging to the replacement target group;
Replacement means for replacing the color of each pixel belonging to the replacement target group with the color of surrounding pixels not belonging to the replacement target group;
Digital image compression means for compressing the digital image after the color of each pixel is replaced by the replacement means to generate compressed image data;
Output means for associating and outputting color data representing a representative color of the replacement target group, the position information, and the compressed image data;
An image compression apparatus comprising: Reading means for reading a document and generating a digital image;
An image compression apparatus according to claim 10;
An image reading apparatus comprising: Setting means for setting a processing target area on the digital image;
Grouping means for grouping pixels constituting the processing target area for each pixel having a small color difference;
Selecting means for selecting a replacement target group based on the feature amount of each of the groups;
Position information generating means for generating position information indicating the position of each pixel belonging to the replacement target group;
Replacement means for replacing the color of each pixel belonging to the replacement target group with the color of surrounding pixels not belonging to the replacement target group;
Digital image compression means for compressing the digital image after the color of each pixel is replaced by the replacement means to generate compressed image data;
An image compression program that causes a computer to function as output means that outputs color data representing a representative color of the replacement target group, the position information, and the compressed image data.

Images