JP2009105541A - Image processing apparatus, method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technique that can appropriately handle characters in various types of images. <P>SOLUTION: A dot area representing a dot image and a character string area, representing a character string and different from the dot area are detected in an object image, and the character string area in the object image is enlarged. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image processing apparatus, a method, and a program.

  Conventionally, techniques for handling characters in an image are known. For example, a technique for classifying and expanding character elements character by character is known.

Japanese Patent Laid-Open No. 11-25283

  By the way, there are images representing various types of objects such as characters, illustrations, and photographs. However, in the past, the actual situation is that no sufficient contrivance has been made to handle characters appropriately in such various images.

  SUMMARY An advantage of some aspects of the invention is to provide a technique capable of appropriately handling characters in various images.

  SUMMARY An advantage of some aspects of the invention is to solve at least a part of the problems described above, and the invention can be implemented as the following forms or application examples.

Application Example 1 An image processing apparatus for processing target image data representing a target image, wherein a halftone dot region representing a halftone dot image and a character representing a character string that is different from the halftone dot region from the target image An image processing apparatus comprising: a region classification unit that detects a row region; and an image processing unit that enlarges the character string region in the target image.

  According to this configuration, a halftone dot region and a character string region different from the halftone dot region are detected and the character string region is enlarged, so that characters can be handled appropriately in various images.

[Application Example 2] The image processing apparatus according to Application Example 1, wherein the area classifying unit applies an area representing the halftone dot image to the halftone dot regardless of whether the halftone dot image includes characters. An image processing apparatus that detects an area.

  According to this configuration, it is possible to prevent the halftone dot region from being excessively deformed due to the influence of the character string expansion regardless of whether or not the halftone dot image includes characters.

Application Example 3 In the image processing apparatus according to Application Example 1 or Application Example 2, the area classification unit detects a blank area, and the image processing unit uses the character string area as the character string area. An image processing apparatus that enlarges within the margin area adjacent to the image processing apparatus.

  According to this configuration, since the character string area is enlarged in the blank area adjacent to the character string area, it is possible to suppress the image layout from being excessively collapsed.

Application Example 4 In the image processing device according to any one of Application Examples 1 to 3, the region classification unit selects a character pixel representing a character from a plurality of pixels included in the target image data. An image processing apparatus that detects a region including a connected region obtained by detecting and connecting character pixels arranged apart from each other as the character string region.

  According to this configuration, since the character string area is detected by connecting the character pixels, the processing load for detecting the character string area can be reduced.

Application Example 5 The image processing apparatus according to Application Example 4, wherein the region classification unit obtains the connected region by anisotropic connection that connects the character pixels along a predetermined direction. .

  According to this configuration, the character string region can be detected appropriately.

[Application Example 6] The image processing apparatus according to Application Example 5, wherein the region classification unit is a result of each of N types (N is an integer of 2 or more) of anisotropic connections in which the predetermined directions are different from each other. An image processing apparatus that determines a representative value of a length along the predetermined direction of an area where the character pixels are connected, and detects the character string area according to the anisotropic connection having the longest representative value.

  According to this configuration, it is possible to appropriately detect the character string region with respect to various target images.

Application Example 7 An image processing method for processing target image data representing a target image, wherein a halftone dot region representing a halftone dot image is different from the halftone dot region and a character representing a character string from the target image An image processing method comprising: detecting a row region; and enlarging the character string region in the target image.

Application Example 8 A computer program for processing target image data representing a target image, wherein a halftone dot region representing a halftone dot image is different from the halftone dot region and represents a character string from the target image A computer program that causes a computer to realize a function of detecting a character string region and a function of expanding the character string region in the target image.

[Application Example 9] The image processing apparatus according to Application Example 2, wherein the image processing unit enlarges the character string area without enlarging the halftone dot area.

  According to this configuration, since the halftone dot region is not enlarged, the layout of the target image can be prevented from being excessively collapsed.

  The present invention can be realized in various forms, for example, an image processing method and apparatus, a computer program for realizing the function of the method or apparatus, a recording medium on which the computer program is recorded, Or the like.

Next, embodiments of the present invention will be described in the following order based on examples.
A. First embodiment:
B. Modified examples of character string area expansion:
C. Variations on string detection:
D. Example of pixel classification processing:
E. Variations:

A. First embodiment:
Examples of the present invention will be described.
A-1. General configuration of the printer 10:
FIG. 1 is an explanatory diagram showing a schematic configuration of a printer 10 as an embodiment of the image processing apparatus of the present application. The printer 10 is a so-called multifunction printer having a scanner function and a copy function in addition to a print function. The printer 10 includes a control unit 20, a carriage moving mechanism 60, a carriage 70, a paper feed mechanism 80, a scanner 91, and an operation panel 96.

  The carriage moving mechanism 60 includes a carriage motor 62, a drive belt 64, and a sliding shaft 66, and drives a carriage 70 that is movably held on the sliding shaft 66 in the main scanning direction. The carriage 70 includes an ink head 71 and an ink cartridge 72, and discharges ink supplied from the ink cartridge 72 to the ink head 71 onto the printing paper P. The paper feed mechanism 80 includes a paper feed roller 82, a paper feed motor 84, and a platen 86. The paper feed motor 84 rotates the paper feed roller 82 to convey the printing paper P along the upper surface of the platen 86. To do. The scanner 91 is an image scanner that optically reads an image. In this embodiment, the CCD (Charge Coupled Devices) method is used, but various methods such as a CIS (Contact Image Sensor) method can be used.

  Each mechanism described above is controlled by the control unit 20. The control unit 20 is configured as a microcomputer including a CPU 30, a RAM 40, and a ROM 50. The program stored in the ROM 50 is expanded and executed on the RAM 40, thereby performing control of each mechanism described above, as shown in FIG. Functions as a functional unit. Details of these functional units will be described later.

  The printer 10 having the above configuration functions as a copier by printing an image read by the scanner 91 on the printing paper P. The above printing mechanism is not limited to the ink jet type, and various printing methods such as a laser type and a thermal transfer type can be used.

A-2. Image duplication processing:
FIG. 2 is a flowchart showing the flow of image duplication processing for copying a predetermined image using the printer 10. FIG. 3 is a schematic diagram showing this image duplication processing. This process is started when the user sets an image to be copied (for example, a document such as a printed material) on the printer 10 and performs a copy instruction operation using the operation panel 96. When this process is started, the CPU 30 converts the formed optical image into an electrical signal using the scanner 91 as an image input process (step S100). Then, as an image conversion process, the obtained analog signal is converted into a digital signal by an AD conversion circuit, and further shading correction is performed so that the entire image has uniform brightness (step S110).

  In the next step S122, the CPU 30 performs region classification on the obtained image data (also referred to as “target image data SI”) (step S122). In this process, the entire target image SII represented by the target image data SI is classified into a halftone dot area representing a halftone dot portion, a character string area representing a character string, and a blank area representing a margin. The classification image DAI in FIG. 3 represents the result of the area classification processing for the target image SII. The target image SII represents five character strings CS1 to CS5 extending in the horizontal direction (x direction) and a picture portion PB. The picture portion PB represents an image obtained by superimposing the character CSP and a photograph. In the original image (for example, a document) to be copied, this picture portion PB is represented by a halftone dot.

  As indicated by the classified image DAI, the target image SII is classified into five character string areas CA1 to CA5, one halftone dot area DA, and one blank area BA. Five character string areas CA1 to CA5 represent five character strings CS1 to CS5, respectively. A halftone dot area DA represents a picture portion PB represented by a halftone dot. The margin area BA represents the remaining part. Details of the area classification process will be described later.

  In the next step S132, the CPU 30 generates an enlarged image in which the character string area is enlarged. In this process, the character string area is enlarged. In the enlarged area image EAI of FIG. 3, the five character string areas CA1 to CA5 in the classified image DAI are replaced with five enlarged character string areas ECA1 to ECA5, respectively. Then, the image represented by the character string region in the target image SII is enlarged so as to coincide with the enlarged character string region. In the enlarged image PI of FIG. 3, the five character strings CS1 to CS5 in the target image data SI are respectively replaced with the five enlarged character strings ECS1 to ECS5. Such an enlarged image PI is generated by synthesizing the target image SII and the enlarged character string image. Details of the process of generating the enlarged image PI (also referred to as “image composition process”) will be described later.

  When the image composition processing is performed, the CPU 30 performs overall correction processing such as gamma correction and color correction that reduces the error in color information between the input image and the output image so that the color tone and the like can be reproduced during output (step S140). As the processing of the print control unit 35, the carriage moving mechanism 60, the carriage 70, the paper feed mechanism 80, etc. are driven to output an image on the print paper P (step S150). In this way, the image duplication process is completed.

Area classification processing:
FIG. 4 is a flowchart showing the procedure of the area classification process shown in step S122 of FIG. In the first step S700, the CPU 30 reads the image data (in this case, RGB data) obtained in step S110 into the RAM 40 as processing of the image input unit 31. And CPU30 performs the process of following step S710-S750 as a process of the area | region classification | category part 33. FIG.

  In step S <b> 710, the CPU 30 performs determination processing for each pixel (also referred to as attribute determination processing) as processing of the region classification unit 33. Here, the attribute of the pixel represents an attribute relating to the type of the image area represented by the pixel. In this embodiment, by this processing, the plurality of pixels representing the target image SII are classified into character pixels representing characters, halftone dots representing halftone dots, and blank pixels representing blanks. As will be described later, when a character is represented by a halftone dot, the pixel representing the character is classified as a halftone pixel. As such a pixel classification method, any method can be adopted. Details of the pixel classification embodiment will be described later.

  In the next steps S 720 and S 730, the CPU 30 detects a character string region as a process of the character string detection unit 336 of the region classification unit 33. First, the CPU 30 connects character pixels by expanding and shrinking a graphic (S720).

  FIG. 5 is a schematic diagram of graphic enlargement and reduction. In the figure, target image data SI is shown. The target image data SI represents each gradation value of a plurality of pixels pix arranged in a matrix along the horizontal direction (x direction) and the vertical direction (y direction). Hereinafter, it is assumed that the pixels pix are arranged in the + x direction and the + y direction with reference to the pixel at the top left (corner) (referred to as “reference pixel pix_s”). In the drawing, a plurality of pixels pix are divided into character pixels and non-character pixels (pixels that are not character pixels).

  The lower part of FIG. 5 shows an outline of the enlargement and reduction of the figure. The enlargement of a graphic is a process of setting the attribute of a pixel around the pixel of interest pix_i to a character pixel when the pixel of interest pix_i is a “character pixel”. As a result, the character (character pixel region) is enlarged. Hereinafter, a pixel area whose attribute is set to “character pixel” is referred to as a “character enlargement area”. In the example of FIG. 5, a predetermined rectangular area centered on the target pixel pix_i is used as the character enlargement area. The rectangular region has a length in the x direction of 9 pixels and a length in the y direction of 3 pixels. As a result, the character pixel region expands by four pixels in each of the + x direction and the −x direction, and further, the character pixel region expands by one pixel in each of the + y direction and the −y direction. Thus, in the embodiment shown in FIG. 5, the characters are widened along the x direction. Note that, in the enlargement of the graphic, the above-described rectangular area is applied to all the pixels of the target image data SI.

  On the other hand, the graphic degeneration is a process of setting the attribute of pixels around the target pixel pix_i to “non-character pixel” when the target pixel pix_i is “non-character pixel”. As a result, the character (character pixel region) is degenerated. Hereinafter, a pixel area whose attribute is set to “non-character pixel” is referred to as a “character degenerate area”. In the example of FIG. 5, the same 9 × 3 pixel rectangular area as the character enlargement area is used as the character degeneration area. As a result, the character pixel area is reduced by four pixels in each of the + x direction and the −x direction, and the character pixel area is reduced by one pixel in each of the + y direction and the −y direction. . As described above, in the embodiment shown in FIG. 5, the character is strongly degenerated along the x direction. In the graphic reduction, the above rectangular area is applied to all the pixels of the target image data SI.

  6 (A) to 6 (C) and FIGS. 7 (A) to (C) are schematic diagrams showing the connection of character pixels by enlargement and reduction of a figure. Each figure shows a plurality of pixels arranged along the x and y directions. One black dot in the figure represents one character pixel. A portion without black dots represents a “non-character pixel”.

  FIG. 6A shows an example of the result of step S710 in FIG. This example represents two character strings CSa and CSb extending in the x direction. In the present embodiment, the CPU 30 repeats the enlargement of the figure twice, and then repeats the reduction of the figure twice. Thereby, the attribute of each pixel changes in the order of FIGS. 6 (A), 6 (B), 6 (C), 7 (A), and 7 (B).

  By the first graphic enlargement, the attribute of each pixel changes from FIG. 6 (A) to FIG. 6 (B). By the next graphic enlargement, the attribute of each pixel changes from FIG. 6 (B) to FIG. 6 (C). Since the area of the character pixel is expanded by the graphic enlargement, the character pixel representing each character is connected by the newly set character pixel. In particular, in this embodiment, the character pixel area is widened along the x direction (FIG. 5). As a result, in each of the two character strings CSa and CSb extending in the x direction, characters are connected along the x direction.

  Next, the attribute of each pixel changes from FIG. 6C to FIG. The attribute of each pixel changes from FIG. 7 (A) to FIG. 7 (B) by the next graphic reduction. Due to these graphic reductions, the excessively expanded character pixel region is reduced. However, since the characters in the character string are filled with character pixels, the connection between the characters by the character pixels is maintained. As a result, the CPU 30 forms a region in which character pixels of a plurality of characters are connected along the x direction.

  In the next step S730 in FIG. 4, the CPU 30 detects a character string area. FIGS. 7B and 7C show the detected character string areas CAa and CAb. In this embodiment, a rectangular area including an area where character pixels are connected is detected as a character string area. One character string area is a minimum rectangular area including one connected area. The four vertices of this rectangular area are set to four pixel positions obtained by combining the maximum pixel position and minimum pixel position in the x direction and the maximum pixel position and minimum pixel position in the y direction in one connected area. Such a character string area is detected for each connected area. In the example of FIG. 7B, two character string areas CAa and CAb are detected. As described above, the CPU 30 detects a continuous character string region in which a plurality of characters are connected.

  FIG. 7C shows a state in which two character string areas CAa and CAb are superimposed on the same image as FIG. As shown in the drawing, the first character string area CAa is a minimum rectangular area that includes character pixels representing the first character string CSa. The second character string area CAb is a minimum rectangular area that includes character pixels representing the second character string CSb. In the present embodiment, the character enlargement area and the character reduction area are the same, and the number of graphic enlargements and the number of figure reductions are also the same. Therefore, it is possible to detect a character string area without excess or deficiency.

  In the embodiment shown in FIG. 5, a process (enlargement / reduction) for widely connecting characters (character pixels) along the x direction is adopted, but the characters are expanded along other directions (for example, the y direction). You may employ | adopt the process to connect. Hereinafter, pixel connection having such directionality is referred to as anisotropic connection.

  In the next step S740 in FIG. 4, the CPU 30 detects a blank area as the process of the blank detection unit 338 of the area classification unit 33. Specifically, the CPU 30 determines an area represented by the remaining pixels obtained by removing the pixels (character pixels) included in the character string area detected in step S730 from the blank pixels specified in step S710. Detect as. From the detected margin area, the size of the margin (number of pixels) around the character string area and the distance between the character string areas (line spacing) can be specified.

  In the next step S750, the CPU 30 writes the result of area classification into the RAM 40 as the process of the area classification unit 33. In this embodiment, data indicating whether each pixel of the target image data SI belongs to a character string area, a halftone dot area, or a margin area is written in the RAM 40. Then, the CPU 30 ends the area classification process and returns the process to the image duplication process of FIG. Note that the CPU 30 (region classification unit 33) directly adopts the region represented by the halftone pixel detected in step S710 as the halftone dot region.

Image composition processing:
FIG. 8 is a flowchart showing the procedure of the image composition process shown in step S132 of FIG. The CPU 30 executes steps S800 to S830 in FIG. 8 as processing of the image composition unit 36. The CPU 30 refers to the target image data SI stored in the RAM 40 in the first step S800, and refers to the result of area classification stored in the RAM 40 in the next step S810.

  In the next step S820, the CPU 30 enlarges the character string area without enlarging the halftone dot area. FIG. 9 is a schematic diagram of character string region expansion. In the figure, the same classified image DAI and enlarged region image EAI as the image shown in FIG. 3 are shown. The CPU 30 enlarges each of the detected one or more character string regions to the maximum within a range that satisfies the conditions described later. The conditions are as follows. The CPU 30 enlarges the character string area within a blank area that is in contact with the character string area. Here, the position of the center of gravity of the character string area is fixed. Further, the aspect ratio of the character string area is not maintained, that is, the enlargement ratio in the x direction and the enlargement ratio in the y direction are adjusted independently of each other. When a plurality of character string areas are arranged with a blank area in between, the character string areas are enlarged so that the character string areas do not overlap. At this time, the enlargement ratio in the x direction and the enlargement ratio in the y direction are common to each character string region. As a result, the five character string areas CA1 to CA5 in the classified image DAI are enlarged at the same enlargement ratio, and five enlarged character string areas ECA1 to ECA5 are respectively generated.

  Further, in this embodiment, the CPU 30 leaves a margin BC around the enlarged character string area. In this way, the enlarged character string is prevented from coming into contact with other image portions (for example, other character strings), so that the character string can be easily read. Various sizes can be adopted as the size of the margin BC (in this case, the size of the margin BC is changed from an enlarged character string area to an image area other than the margin area (for example, another extended character string area). Corresponds to the shortest distance to)). For example, the CPU 30 can employ a predetermined size (for example, a predetermined number of pixels or a predetermined length in a print result). Instead, the CPU 30 may determine the size of the margin BC according to the target image data SI. For example, the size of the margin BC around the enlarged character string region may be increased as the width WC (length in the direction perpendicular to the longitudinal direction) of the enlarged character string region is increased. Further, the CPU 30 may determine the size of the margin BC according to a user instruction. Note that the CPU 30 may enlarge the character string area to the maximum without leaving a margin BC around the enlarged character string area.

  In the present embodiment, the CPU 30 leaves a margin BE on the edge ED of the enlarged region image EAI (target image SII). By doing so, the enlarged character string is prevented from coming into contact with the edge ED, so that the character string can be easily read. Various sizes can be adopted as the size of the margin BE (in this case, the size of the margin BE is the shortest distance from the edge ED of the enlarged region image EAI (target image SII) to the enlarged character string region. Equivalent to). For example, the CPU 30 can employ a predetermined size (for example, a predetermined number of pixels or a predetermined length in a print result). Instead, the CPU 30 may determine the size of the margin BE according to the print setting. For example, some printers can perform borderless printing without leaving margins on the four sides of the printing paper P. In borderless printing, printing data is slightly enlarged and printed without margins. Here, the size of the margin BE may be determined so that the enlarged character string region does not protrude from the printing paper P. Here, the size of the margin BE may be a variable value according to the size of the printing paper P. The CPU 30 may enlarge the character string area to the maximum without leaving the margin BE on the edge ED.

  Next, the CPU 30 enlarges the image in the character string area in accordance with the enlarged character string area. The size, shape, and position of the enlarged image are the same as the enlarged character string region associated with the character string region. Various methods can be adopted as the image enlargement method. For example, affine transformation (linear transformation) may be employed. In the embodiment shown in FIG. 3, the CPU 30 enlarges the image representing the character string CS1 to generate an image representing the enlarged character string ECS1. The same applies to the other character strings CS2 to CS5. Then, the CPU 30 generates an enlarged image PI (generates image data representing the enlarged image PI) by synthesizing the generated image and the target image SII. In this embodiment, the CPU 30 does not enlarge the halftone dot area. As a result, the character CSP represented by the halftone dots is not enlarged.

  In the next step S830 in FIG. 8, the CPU 30 stores image data representing the enlarged image PI in the RAM 40. Then, the CPU 30 ends the image composition process and returns the process to the image duplication process of FIG. In steps S140 and S150 of FIG. 2, processing is executed according to the image data representing the enlarged image PI.

  As described above, in this embodiment, since the character string in the target image SII is enlarged, the character string can be easily read without enlarging the entire image. Moreover, since the continuous character string area | region showing a character string (a several character) is expanded, it can suppress that a character collapses in the character string after expansion. Further, since the line representing the character becomes thick by enlarging the image representing the character string, the readability of the character can be improved. Further, since the character string region is enlarged within the margin in the target image SII, it is possible to suppress the layout of the enlarged image PI from being excessively collapsed with respect to various target images SII. As a result, it is possible to prevent the user who has viewed the enlarged image PI from misreading the character string. Furthermore, since the halftone dot region is not included in the enlargement target, it is possible to suppress the layout of the enlarged image PI from being excessively collapsed due to the enlargement of an image (for example, a photograph or an illustration) represented by the halftone dot. And it can suppress that it becomes impossible to expand a character string. In this embodiment, an area representing a halftone dot image is detected as a halftone dot area regardless of whether the halftone dot image includes characters. Therefore, even when a character is included in the image portion represented by the halftone dots, the character portion is prevented from being used as an enlargement target. As a result, the image portion represented by the halftone dot can be prevented from being unnaturally deformed.

  Moreover, as shown in FIGS. 5-7, since the area | region containing the connection area | region obtained by connecting between the character pixels arrange | positioned apart is used as a character string area | region, one character string area | region is object. It is possible to include the entire character string in the image data SI. Since one character string region is enlarged without being divided, it is possible to suppress the format from being changed in the middle of the character string (for example, it is possible to suppress a line break in the middle of the character string). As a result, it is possible to prevent the user who has viewed the enlarged image PI from misreading the character string. Moreover, since the analysis for recognizing characters one by one can be omitted, the processing load can be reduced.

  Moreover, as shown in FIGS. 5-7, since a character pixel is connected by anisotropic connection, a character string area | region can be detected appropriately. Moreover, it can suppress that the part showing a line space is recognized as an area | region showing a character. As a result, line spacing can be used to expand the character string area, and it is easy to enlarge the characters.

  In this embodiment, the image composition unit 36 (FIG. 1) corresponds to an “image processing unit” in the claims.

B. Modified examples of character string area expansion:
FIG. 10 is a schematic diagram of another embodiment of character area expansion. The only difference from the embodiment shown in FIG. 9 is that the barycentric position of the character area is variable. FIG. 10 shows a classified image DAIa and an enlarged area image EAIa. The classification image DAIa is the same as that obtained by deleting the fourth character string area CA4 and the fifth character string area CA5 from the classification image DAI of FIG. Therefore, the blank area BA between the halftone dot area DA and the character string area is wider than the classified image DAI of FIG.

  The CPU 30 (FIG. 1) generates three enlarged character string areas ECA1a to ECA3a by enlarging the three character string areas CA1 to CA3, respectively. At this time, the CPU 30 makes maximum use of the blank area BA between the character string area and the halftone dot area DA. Specifically, the CPU 30 moves the gravity center positions of the enlarged character string areas ECA1a to ECA3a to the halftone dot area DA side from the gravity center positions of the character string areas CA1 to CA3. As a result, the enlarged character string areas ECA1a to ECA3a can be enlarged as shown in the enlarged area image EAIa of FIG.

  It should be noted that various other modes can be employed for expanding the character string area. For example, the aspect ratio of the character string area may be maintained. In this way, it is possible to suppress excessive deformation of the characters. Further, the upper limit of the enlargement rate of the character string area may be determined in advance. In this way, it is possible to suppress an excessively large character string. Further, when a plurality of character string areas are arranged with a blank area in between, the enlargement ratio may be different for each character string area. In addition, the enlarged character string area may extend in an area other than the margin area (for example, a halftone dot area).

C. Variations on string detection:
FIG. 11 is a schematic diagram showing another embodiment of character string detection. The example of FIG. 11 shows processing executed instead of step S720 of FIG. The difference from the embodiment shown in FIG. 5 is that, instead of connecting character pixels along a predetermined direction, a result of connecting character pixels widely along the x direction and a result of connecting character pixels widely along the y direction Only one point is selected by comparing these results. By this selection, it is possible to detect a character string suitable for the direction in which the character string extends in the target image data SI.

  In the first step S900, the CPU 30 executes enlargement / reduction to widely connect characters (character pixels) along the x direction with respect to the character pixels of the target image data SI (hereinafter referred to as “x direction connection”). ). This x-direction connection is the same as the processing described in FIGS. Then, the CPU 30 generates an x-direction coupled image CI_x representing the result of the x-direction coupling. The x-direction connected image CI_x represents a connected region in which characters (character pixels) are connected, as in the image of FIG. 7B. The longest connected area CAx_max in the x-direction connected image CI_x indicates a connected area having a maximum length along the x direction (this length is referred to as “x-direction maximum length Lx_max”).

  In the next step S910, the CPU 30 executes enlargement / reduction to widely connect characters (character pixels) along the y direction with respect to the character pixels of the target image data SI (hereinafter referred to as “y direction connection”). ). In this y-direction connection, a rectangular area of 3 (x direction) × 9 (y direction) pixels is used instead of the rectangular area of 9 × 3 pixels as the character enlargement area and the character contraction area. Thereby, a character (character pixel) is widely connected along the y direction. Then, the CPU 30 generates a y-direction connected image CI_y representing the result of the y-direction connection. The y-direction connected image CI_y also represents a connected area in which characters (character pixels) are connected. The longest connected area CAy_max in the y-direction connected image CI_y indicates a connected area having a maximum length along the y direction (this length is referred to as “y-direction maximum length Ly_max”).

  In the next step S920, the CPU 30 compares the maximum x-direction length Lx_max with the maximum y-direction length Ly_max. When the x-direction maximum length Lx_max is larger than the y-direction maximum length Ly_max, the CPU 30 selects the x-direction connected image CI_x. Then, in the next step S730 in FIG. 4, the CPU 30 detects a character string area according to the selected x-direction connected image CI_x. Conversely, when the x-direction maximum length Lx_max is equal to or less than the y-direction maximum length Ly_max, the CPU 30 selects the y-direction connected image CI_y. Then, in the next step S730 in FIG. 4, the CPU 30 detects a character string region according to the selected y-direction connected image CI_y.

  Thus, the reason for selecting the longer one of the maximum lengths of the x direction connection and the y direction connection is as follows. That is, the extending direction of the character string in the target image data SI can take various directions. For example, the character string may extend in the x direction or the character string may extend in the y direction. In either case, the line spacing is usually wider than the character spacing. Therefore, for a target image in which a character string extends in the x direction, a connection region in which the x direction connection is longer than the y direction connection is obtained. The x-direction connected image CI_x and the y-direction connected image CI_y shown in FIG. 11 show such a case. On the contrary, for a target image in which a character string extends in the y direction, a connected region in which the y direction connection is longer than the x direction connection is obtained. As described above, it is possible to detect a character string suitable for the direction in which the character string extends in the target image data SI by selecting the longer one of the x-direction connection and the y-direction connection.

  The comparison target is not limited to the maximum length, and a representative value obtained from the lengths of one or more connected regions included in the result of connecting character pixels can be used. As the representative value, various values represented by a function of the length of each connected region can be adopted (for example, an average value, a maximum value, a median value, a minimum value). In any case, in order to evaluate the result of widely connecting character pixels along a specific direction (result of anisotropic connection), the representative value of the length of the connected region along that direction can be used. Good. And it can be said that the larger the representative value is, the more suitable the anisotropic connection is for the target image data SI.

  In addition, the choice of the anisotropic coupling direction is not limited to the two directions of the x direction and the y direction perpendicular to the x direction, and various directions can be adopted. For example, three directions including an x direction, a y direction, and a direction that forms an angle of 45 degrees with the x direction may be employed. In general, the character string detection unit 336 (CPU 30) can use N types of anisotropic connections (N is an integer of 2 or more) having different directions. Here, the character string detection unit 336 (CPU 30) may detect the character string region in accordance with the anisotropic connection having the longest representative value. Instead, the character string detection unit 336 (CPU 30) may select an anisotropic connection to be used from among N types of anisotropic connections in accordance with a user instruction.

  Further, the process of widely connecting character pixels along a specific direction (anisotropic connection process) is not limited to the process shown in FIG. 5, and various processes can be employed. For example, each of the character enlargement region and the character reduction region is not limited to a 9 × 3 pixel rectangular region, and various anisotropic regions (for example, a 10 × 3 rectangular region or an elliptical region) can be employed. is there. The character contraction area may be different from the character enlargement area. The number of times of figure enlargement and figure reduction can be arbitrarily set, and the number of figure reductions may be different from the number of figure enlargements. Moreover, you may abbreviate | omit figure reduction. In general, as an anisotropic connection process, a distance that can be connected between two character pixels that are separated by one or more pixels (referred to as “connectable distance”) is a connectable distance along a specific direction. Arbitrary processing that is longer than the connectable distance along the direction perpendicular to the specific direction can be adopted.

  Further, the process of connecting the character pixels is not limited to the process having such directionality, and other processes can be employed. For example, in the processing shown in FIG. 5, the character enlargement region and the character degeneration region may be isotropic regions (substantially circles) or may be squares.

D. Example of pixel classification processing:
D-1. First example of pixel classification processing:
FIG. 12 is a flowchart showing the flow of the pixel classification process (area classification process for each pixel) shown in step S710 of FIG. When this process is started, the CPU 30 first refers to the target image data SI stored in the RAM 40 (S200).

  In the next step S205, the CPU 30 generates the first integration data ID1 (FIG. 1) and the second integration data ID2 (FIG. 1) by analyzing the target image data as the processing of the integration data generation unit 32. When the image data is read in units of bands, the integrated data ID1 and ID2 may be generated in units of bands. The CPU 30 stores the generated integration data ID1 and ID2 in the RAM 40. Details of the integral data ID1 and ID2 will be described later in “D-2. Calculation Using Integral Data”.

  In the next step S210, the CPU 30 calculates a statistical variance (hereinafter simply referred to as “dispersion”) of luminance values in the partial region including the target pixel as the processing of the feature amount calculation unit 332 included in the region classification unit 33. To do. FIG. 13 is a schematic diagram illustrating target image data SI and a target pixel pix_k. Note that the luminance value represented by the pixel can be obtained by a known method from the RGB gradation values represented by the pixel.

  In FIG. 13, the pixel of interest pix_k is shown. The target pixel pix_k is a target pixel for attribute determination as to whether the pixel corresponds to an edge portion or a halftone dot portion. In the present embodiment, the CPU 30 determines the attribute of each pixel by sequentially moving the target pixel from the upper left corner pixel of the image to the lower right corner pixel.

  As will be described later, for determination of the pixel of interest pix_k, luminance value dispersion in each of the three partial regions SAk0, SAk1, and SAk2 including the pixel of interest pix_k is used. Each region SAk0 to SAk2 is a square region centered on the pixel of interest pix_k (the first partial region SAk0 corresponds to 5 × 5 pixels, the second partial region SAk1 corresponds to 7 × 7 pixels, The partial area SAk2 corresponds to 9 × 9 pixels).

  The CPU 30 calculates the variance of the partial areas SAk0, SAk1, and SAk2 according to the following formula 1.

As will be described later, the CPU 30 calculates the average value E (f) of the luminance value f by using the first integral data ID1 (FIG. 1), and uses the second integral data ID2 to calculate the luminance value f. An average value E (f ² ) of squares is calculated. The calculated variance is used in the next step S220.

  Next, the CPU 30 performs region determination processing (also referred to as attribute determination processing) as processing of the region determination unit 334 (also referred to as attribute determination unit 334) of the region classification unit 33, and whether or not the target pixel is an edge constituent part. It is determined whether it is a halftone dot component (step S220). Details of the attribute determination process will be described later in “D-3. Attribute Determination Process”.

  In step S220, the CPU 30 first classifies the plurality of pixels of the image data SI into halftone pixels and edge pixels. Next, the CPU 30 identifies a pixel satisfying a predetermined condition among the halftone pixels as a blank pixel. As the blank pixel condition, any condition indicating that the background portion of the target image SII is represented can be employed. For example, it is possible to employ a color within a predetermined color range (for example, a color range having a luminance value equal to or greater than a predetermined threshold). Next, the CPU 30 identifies a pixel in the closed region surrounded by the edge pixels among the plurality of pixels of the target image data SI as a character pixel. As a result, the CPU 30 classifies the plurality of pixels of the target image data SI into character pixels, halftone pixels, and blank pixels.

  When determining the attribute of the pixel of interest, the CPU 30 writes the result in the RAM 40 (step S240). Then, it is determined whether or not the above process has been completed for all the pixels of the target image represented by the target image data (step S250), and if not (step S250: NO), the process is performed in the above step. Returning to S210, if completed (step S250: YES), the pixel classification process is terminated, and the process returns to the area classification process of FIG.

D-2. Calculation using integral data:
FIG. 14 is a schematic diagram of the target image data SI and the first integration data ID1. Hereinafter, the xa-th pixel in the x direction and the ya-th pixel in the y direction from the reference pixel pix_s are represented by pix (xa, ya). Further, the luminance value of the pixel is represented by f (xa, ya). Note that the range of xa is 1 to Nx, and the range of ya is 1 to Ny (Nx and Ny are integers and are determined according to target image data SI).

  The first integral data ID1 represents the luminance integral value (total luminance value) of each pixel pix of the target image data SI. The luminance integrated value p (xt, yt) of a certain pixel pix (xt, yt) is expressed by the following formula 2.

  This luminance integral value p (xt, yt) represents the total value of the luminance values in the rectangular area IAt in which the two pixels of the reference pixel pix_s and the pixel pix (xt, yt) are diagonal pixels ( In FIG. 14, the rectangular area IAt is hatched).

  FIG. 15 is a schematic diagram showing average value calculation using the first integral data ID1. FIG. 15 shows the same target image data SI and first integrated data ID1 as in FIG. Each image data SI and ID1 shows a target rectangular area Ad centered on the pixel of interest pix_k (= pix (xk, yk)). The number of pixels W in the x direction of the target rectangular area Ad is “2 × dx + 1”, and the number of pixels H in the y direction is “2 × dy + 1” (dx is an integer of 1 or more, dy is an integer of 1 or more) .

  In the drawing, four corner pixels pix_a to pix_d of the target rectangular area Ad are shown. The first corner pixel pix_a is the corner pixel closest to the reference pixel pix_s. The fourth corner pixel pix_d is a corner pixel farthest from the reference pixel pix_s. The second corner pixel pix_b is another corner pixel included in the same pixel row as the first corner pixel pix_a, and the third corner pixel pix_c is another corner pixel included in the same pixel column as the first corner pixel pix_a. is there.

  Further, FIG. 15 shows a corner rectangular area AL. This corner rectangular area AL is a rectangular area in which the two pixels of the reference pixel pix_s and the fourth corner pixel pix_d are diagonal pixels. This corner rectangular area AL is divided into four areas Aa, Ab, Ac, Ad by two orthogonal straight lines L1, L2. The first straight line L1 is a straight line parallel to the x direction passing through the −y side edge (pixel boundary line representing the boundary line between adjacent pixels) of the target rectangular area Ad, and the second straight line L2 is the target rectangular area It is a straight line parallel to the y direction passing through the side (pixel boundary line) on the −x side of Ad. The first area Aa is a rectangular area that includes the reference pixel pix_s and is located on the upper left (−x and −y side) in the diagonal direction of the target rectangular area Ad. The second area Ab is a rectangular area located on the −y side of the target rectangular area Ad. The third area Ac is a rectangular area located on the −x side of the target rectangular area Ad.

  The upper part of FIG. 15 shows a comparative example for calculating the average value E (f). The average value E (f) represents the average value of the luminance values f in the target rectangular area Ad. In the comparative example, the luminance value of each pixel in the target rectangular area Ad is read from the target image data SI, and the total value is calculated. Then, the average value E (f) is calculated by dividing the total value by the total number of pixels (W × H). In this case, it is necessary to access the target image data SI (RAM 40) H × W times in order to calculate the average value E (f) for one pixel of interest pix_k. For example, when H = 9 and W = 9, 81 accesses are required. And calculation (addition) of 81 values is also required.

  The lower part of FIG. 15 shows calculation of the average value E (f) in this embodiment. In the figure, four calculation pixels pix_1 to pix_4 used for calculating the average value E (f) are shown. The CPU 30 selects these four calculation pixels pix_1 to pix_4 according to the positions of the four corner pixels pix_a to pix_d.

  The first calculation pixel pix_1 is a pixel at a position where the row (position in the y direction) and the column (position in the x direction) are shifted one by one to the reference pixel pix_s side from the first corner pixel pix_a. The luminance integral value p (x1, y1) of the first calculation pixel pix_1 represents the total value of the luminance values in the first area Aa.

  The second calculation pixel pix_2 is a pixel at a position where the row (position in the y direction) is shifted by 1 to the reference pixel pix_s side from the second corner pixel pix_b. The luminance integration value p (x2, y2) of the second calculation pixel pix_2 represents the total luminance value of the entire first area Aa and second area Ab.

  The third calculated pixel pix_3 is a pixel at a position where the column (position in the x direction) is shifted by 1 toward the reference pixel pix_s from the third corner pixel pix_c. The luminance integrated value p (x3, y3) of the third calculation pixel pix_3 represents the total value of the luminance values in the first area Aa and the third area Ac.

  The fourth calculation pixel pix_4 is the same as the fourth corner pixel pix_d. The luminance integrated value p (x4, y4) of the fourth calculation pixel pix_4 represents the total luminance value in the entire corner rectangular area AL.

  These four calculated pixels pix_1 to pix_4 are adjacent to the four vertices on the outline of the target rectangular area Ad (rectangular pixel boundary line surrounding the target rectangular area Ad) in the direction of the reference pixel pix_s. . Then, the CPU 30 uses these calculated pixels pix_1 to pix_4 to calculate an average value E (f) according to the following Equation 3.

  In this embodiment, regardless of the sizes of H and W, the average value E (f) for one pixel of interest pix_k can be calculated by four accesses to the first integration data ID1 (RAM 40). . Thus, in the embodiment, the number of accesses to the RAM 40 can be significantly reduced as compared with the comparative example described above. In addition, since the numerator can be calculated by calculating (adding) four values, the calculation load can be greatly reduced.

Calculation of the square of the luminance value f (f ²⁾ of the average value E (f ²⁾ is also similar to the calculation of the average value E of the luminance value f (f). The second integral data ID2 (FIG. 1) generated in step S205 of FIG. 12 represents the luminance square integral value ps of each pixel pix of the target image data SI. The luminance square integral value ps (xt, yt) of a certain pixel pix (xt, yt) is expressed by the following Equation 4.

In step S210 of FIG. 12, the CPU 30 acquires the luminance square integral value ps of the four calculation pixels pix_1 to pix_4 described above from the second integral data ID2 (RAM 40), and squares the luminance value f according to the following equation 5. An average value E (f ² ) is calculated.

In step S210 of FIG. 12, the CPU 30 calculates the variance V (f) using the calculated average values E (f) and E (f ² ) (Formula 1). In the calculation of the variance VAR0 of the first partial area SAk0 (FIG. 13), dx = dy = 2 (FIG. 15). In the calculation of the variance VAR1 of the second partial area SAk1, dx = dy = 3. In the calculation of the variance VAR2 of the third partial area SAk2, dx = dy = 4.

D-3. Attribute judgment processing:
FIG. 16 is a flowchart showing the flow of the area determination process (attribute determination process) shown in step S220 of FIG. The CPU 30 executes the processes of steps S500 to S520 shown in FIG. 16 as a process of the area determination unit 334 included in the area classification unit 33. Such attribute determination processing is also a kind of image processing.

  In the first step S500, the CPU 30 determines whether or not the pixel of interest represents a halftone dot portion using the distribution of luminance values.

  FIG. 17 shows an example of luminance values in two partial regions, that is, an edge portion and a halftone dot portion. When the pixel of interest pix_k represents an edge portion, the color changes greatly in the partial region, and thus the luminance value dispersion is usually large. When the pixel of interest pix_k represents a halftone dot portion, the color changes periodically within the partial region. However, since the color change at the halftone portion is often smaller than that at the edge portion, the dispersion of luminance values is usually smaller than that at the edge portion.

  FIGS. 18A, 18B, and 18C show examples of dispersion histograms in partial regions of 5 × 5 pixels, 7 × 7 pixels, and 9 × 9 pixels, respectively (the horizontal axis is Standard deviation (positive square root of variance) is set to be evenly aligned). Each histogram shows the variance of the halftone dot portion and the edge portion. These histograms are created according to the analysis results of various regions of various image data. As shown in the figure, the deviation of the dispersion distribution is different between the halftone dot portion and the edge portion. Therefore, the edge portion and the halftone dot portion can be determined with a certain degree of accuracy by determining the magnitude of the variance using the threshold value.

  Furthermore, the accuracy of determination can be improved by integrating the dispersion of the plurality of partial regions. FIG. 19 shows an example of a partial region representing a halftone dot portion. In the figure, distributed VARx0 to VARx2 of 5 × 5 pixels, 7 × 7 pixels, and 9 × 9 pixels are shown. The variances VARx0 to VARx2 shown in each histogram of FIG. 18 indicate these variances shown in FIG.

  As shown in FIG. 18, the variance VARx0 of 5 × 5 pixels shows a particularly large value (a value with a relatively low frequency) in a typical variance distribution of the halftone dot portion. As a result, when only this distributed VARx0 is used, there is a high possibility that the target pixel pix_k is erroneously determined to be an edge portion. On the other hand, the variance VARx1 of 7 × 7 pixels and the variance VARx2 of 9 × 9 pixels indicate values close to peaks (values with a relatively high frequency) in a typical variance distribution. As a result, if these distributed VARx1 and VARx2 are used, the possibility of erroneous determination can be reduced.

  The variance of each of the plurality of partial areas changes according to the color change pattern (for example, the subject or halftone dot size represented by the target image) around the pixel of interest pix_k. As a result, the partial area suitable for determination among the plurality of partial areas can change according to the color change pattern. For example, it may be possible to reduce the possibility of erroneous determination by using 5 × 5 pixel dispersion rather than 9 × 9 pixel dispersion.

  Therefore, in this embodiment, in order to increase the determination accuracy regardless of the color change pattern, the CPU 30 determines the respective variances VAR0, VAR1, and VAR2 of the three partial areas SAk0, SAk1, and SAk2 (FIG. 13) having different sizes. Make a judgment using it. Specifically, the CPU 30 determines that the target pixel pix_k is an edge portion when the evaluation value EVa calculated according to the following Equation 6 is 0 or more, and when the evaluation value EVa is less than 0, the halftone dot It is determined to be a part.

  The identifier t is a partial area identifier. In this embodiment, t = 0 corresponds to 5 × 5 pixels, t = 1 corresponds to 7 × 7 pixels, and t = 2 corresponds to 9 × 9 pixels. The maximum value T is the maximum value of the identifier t (2 in this embodiment). The coefficient Cat is a predetermined positive value representing the weight for each partial region. The threshold value THat is a dispersion threshold value determined in advance for each partial region. The variance VARt is the variance in each partial area. The sign function sign is a function that returns the sign of the argument. When VARt> THat, sign = + 1, when VARt = THat, sign = 0, and when VARt <THat, sign = -1.

  The evaluation value EVa represents a value obtained by adding the determination results as to whether or not the variance VARt is larger than the threshold value THat with weighting for each partial region. Accordingly, even when a correct determination result cannot be obtained from a partial area as in the example shown in FIG. 19, a correct determination result can be obtained by using another partial area. As a result, the determination accuracy can be increased regardless of the color change pattern. Note that the threshold values THa0, THa1, THa2 and the coefficients Ca0, Ca1, Ca2 are obtained in advance experimentally and empirically by analyzing a large number of images.

  The CPU 30 calculates the evaluation value EVa and determines whether or not the evaluation value EVa is less than 0 (FIG. 16: S500). If the evaluation value EVa is less than 0, the CPU 30 determines that the pixel of interest is a halftone dot portion (S510), and if the evaluation value EVa is 0 or more, determines that the pixel of interest is an edge portion (S520). And CPU30 complete | finishes an attribute determination process, and returns a process to the pixel classification | category process shown in FIG.

  As described above, the CPU 30 executes classification of character pixels, halftone pixels, and blank pixels according to the classification result of halftone pixels and edge pixels. Normally, the variance of the luminance value of the halftone dot portion is smaller than the variance of the luminance value of the edge portion regardless of whether or not the halftone dot portion includes characters. Therefore, regardless of whether or not the halftone dot portion includes characters, the pixels of the halftone dot portion are classified as halftone dot pixels.

  As described above, in the first example of the pixel classification process, the attribute is determined using the variance of each of the partial areas having different sizes. That is, the attribute is determined in consideration of both a local viewpoint and a global viewpoint. As a result, the determination accuracy can be increased. Further, since the variance is calculated using the integral data, it is possible to suppress an excessive increase in the number of accesses to the memory even when calculating the variance for each of the plurality of pixel positions of the target image data SI. . And calculation load can be reduced.

  In this embodiment, since the change is easy to detect, the luminance gradation value is calculated from the RGB gradation values, and the attribute determination is performed using the calculated luminance gradation value. The value is not limited to luminance, but may be a gradation value representing a color. For example, when the image data is given in the YCbCr format, the Y component of the image data may be used directly as the luminance gradation value, the Cb component or the Cr component may be used, the R component of the pixel, etc. May be used.

  The printer 10 having such a configuration calculates the variance of a pixel group within a predetermined range around the pixel of interest, calculates the evaluation value EVa based on the variance, and determines the attribute of the pixel of interest based on the result. Therefore, it is possible to determine the attributes of pixels such as an edge component and a halftone dot component by simple arithmetic processing centering on distributed calculation. Further, since the determination technique is configured by simple arithmetic processing, it can be configured at low cost by software. Further, since it is configured by a combination of simple arithmetic processing, it can be implemented as SIMD (Single Instruction Multiple Data) parallel processing, and high-speed processing is possible. For example, reading of image data and area classification (determination) may be performed in parallel.

D-4. Second embodiment of pixel classification processing:
FIG. 20 is a flowchart showing another embodiment of the area determination process (attribute determination process) shown in step S220 of FIG. The difference from the embodiment of FIG. 16 is that four types of determination are made for the inside of the character and others in addition to the halftone dot portion and the edge portion. The CPU 30 (FIG. 1) executes the processes of steps S400 to S445 illustrated in FIG. 20 as the processes of the area determination unit 334 included in the area classification unit 33.

  In the first step S400, the CPU 30 determines whether or not the color of the target pixel is included in the background color range. When the color of the target pixel is included in the background color range, the CPU 30 determines that the target pixel is “others” (S445). The background color range indicates a color range representing the background portion of the target image. For example, the CPU 30 employs a color range of a predetermined size centered on the average color of a predetermined edge portion (for example, a portion within 20 pixels from the edge) in the target image as the background color range. For example, when the target image represents a printed matter using white paper, the color range representing the white color of the paper is employed as the background color range. Instead, a predetermined color range (for example, a color range having a luminance value equal to or greater than a predetermined threshold) may be adopted as the background color range.

  In the next steps S410, S420, and S430, the CPU 30 determines whether the target pixel corresponds to the inside of a character, a halftone dot portion, or an edge portion by using the distribution of luminance values.

  FIG. 21 shows an example of luminance values in three partial areas. The only difference from FIG. 17 is that the inside of the character is added. When the pixel of interest pix_k represents the inside of a character, the color of each pixel in the partial area is often almost the same, so that the dispersion of luminance values is usually smaller than that of the halftone dot portion.

  22 (A), 22 (B), and 22 (C) show examples of dispersion histograms in three partial regions, respectively. The only difference from FIG. 18 is that a histogram inside the characters is added. As shown in the figure, the deviation of the dispersion distribution is different between the inside of the character and the halftone dot portion. Therefore, as in the determination of the halftone dot portion and the edge portion, the inside of the character and the halftone dot portion can be determined with a certain degree of accuracy by determining the magnitude of the dispersion with the threshold values THb0, THb1, and THb2.

  In step S410 of FIG. 20, the CPU 30 (FIG. 1) determines whether or not the target pixel is “inside the character” by the same method as in FIG. Step S420 is the same as step S500 in FIG. 16 (determination as to whether or not the pixel of interest is a “halftone dot portion”). In step S430, the CPU 30 determines whether or not the target pixel is an “edge portion” by the same method as in FIG. In step S430, when the variance is excessively large, it is determined that the target pixel is not an edge portion. However, step S430 may be omitted. In this case, all of the pixels that are determined not to be in the background, in the character, or in the halftone portion are determined to be edge portions.

  For each determination in steps S410 and S430, an evaluation value similar to the above-described evaluation value EVa (Formula 6) is used. In step S410, the CPU 30 determines that the target pixel is inside the character when the evaluation value is less than 0, and determines that it is not inside the character when the evaluation value is 0 or more. In step S430, the CPU 30 determines that the pixel of interest is an edge portion when the evaluation value is less than 0, and determines that it is not an edge portion when the evaluation value is 0 or more. A coefficient (corresponding to the coefficient Cat) and a threshold value (corresponding to the threshold value THat) used for calculating each evaluation value are obtained experimentally and empirically in advance by analyzing a large number of images.

  FIG. 23 is a schematic diagram of the determination result of the image portion IP representing characters and halftone dots. The first pattern P1 is a binary pattern indicating whether each pixel is an edge portion. The second pattern P2 is a binary pattern indicating whether each pixel is inside a character. The third pattern P3 is a binary pattern indicating whether each pixel is a halftone dot portion. The fourth pattern P4 is a binary pattern indicating whether each pixel is other. In the figure, pixels corresponding to each portion are represented by solid lines or hatching. As shown in the figure, the pixel representing the outline of the character is determined to be an edge portion, the pixel representing the inside of the character is determined to be inside the character, the pixel representing the halftone image is determined to be a halftone portion, and the background is determined to be others. Is done.

  In step S <b> 240 of FIG. 12, the CPU 30 writes the determination result in the RAM 40 as processing of the region determination unit 334. As the determination result to be written, for example, four binary pattern data shown in FIG. 23 may be adopted, or quaternary pattern data indicating which of the four attributes each pixel is adopted may be adopted. Good.

  Here, the CPU 30 identifies both the edge portion and the inside of the character as character pixels, and identifies the pixels determined to be other as blank pixels. In step S240 in FIG. 12, the CPU 30 may store arbitrary data that can identify the character pixel, the halftone dot pixel, and the blank pixel in the RAM 40. As such data, for example, ternary pattern data indicating which of the three attributes can be employed.

  By combining each embodiment of pixel classification processing as described above and detection of a character string region by character pixel concatenation, the burden required for detection of the character string region can be reduced. For example, it is possible to omit analysis processing for recognizing characters one by one (for example, so-called OCR (optical character recognition) and processing for classifying regions representing characters one by one.

  In addition, as a classification method of the image type which a pixel represents, not only the method of each above-mentioned Example but a well-known various method is employable. Here, if a method of specifying the type of the pixel of interest in accordance with the distribution of pixel values in a partial region including the pixel of interest and the pixels within a predetermined range around the pixel of interest is used, the burden of classification processing can be reduced. it can. For example, in each of the embodiments described above, the maximum difference between the pixel values around the pixel of interest pix_k may be used instead of the variance.

  Further, the types of pixels to be identified are not limited to the three types of margins, halftone dots, and characters, and any type including at least halftone dots and characters can be employed. For example, when the character region extends in other regions (for example, halftone dot regions) other than the margin, the specification of the margin pixel may be omitted.

C. Variations:
In addition, elements other than the elements claimed in the independent claims among the constituent elements in each of the above embodiments are additional elements and can be omitted as appropriate. The present invention is not limited to the above-described examples and embodiments, and can be implemented in various modes without departing from the gist thereof. For example, the following modifications are possible.

Modification 1:
In each of the embodiments described above, a part representing an image different from a character string may be classified as a character string region. For example, when the target image data SI includes a portion representing a graph, the portion representing the graph can be classified as a character string region. In this case, the graph is enlarged similarly to the character string, but there is no practical problem. However, the area representing the graph may be distinguished from the character string area, and the graph area may be excluded from the enlargement target.

Modification 2:
In each of the above-described embodiments, the method for detecting a character string region is not limited to a method combining character pixel detection and character pixel concatenation, and any method can be employed. For example, a character string region may be detected using so-called OCR (optical character recognition).

  Also, various methods can be adopted as a method for detecting a halftone dot region. For example, a halftone dot region may be detected according to pattern matching.

  In any case, it is preferable that the CPU 30 (region classification unit 33) detects a region representing a halftone dot image as a halftone dot region regardless of whether or not the halftone dot image includes characters. In this way, even when characters are included in the image represented by the halftone dots, it is possible to prevent the halftone dot region from being excessively deformed due to the influence of the character string expansion. For example, first, a halftone area is classified (detected) from the target image, and then a character string area is classified (detected) from the remaining area excluding the halftone area.

Modification 3:
In each of the above-described embodiments, the CPU 30 (image composition unit 36) may enlarge the halftone dot area in the same manner as the character string area. Also in this case, the CPU 30 (region classification unit 33) can classify the halftone dot region and the character string region, thereby suppressing the image represented by the halftone dot from being unnaturally deformed.

Modification 4:
In each of the above-described embodiments, the target image data is not limited to the image data generated by scanning by the scanner 91, and various image data can be employed. For example, the printer 10 may acquire target image data from a removable memory card or another device connected via a network.

  In each of the above-described embodiments, the application of the image data after the character string area is enlarged is not limited to printing, and various applications can be employed. For example, an image may be displayed on a display device, and a data file including image data may be provided to the user.

Modification 5:
As mentioned above, although the Example of this invention was described, this invention is not limited to such an Example, Of course, in the range which does not deviate from the summary of this invention, it can implement in a various aspect. For example, the image processing apparatus of the present invention can be mounted not only on the printer multifunction device shown in the embodiment but also on various digital devices such as a printer alone, a digital copying machine, and an image scanner. Further, the present invention is not limited to the configuration as the image processing apparatus, and can be realized in the form of a determination image processing method, a computer program, or the like.

  Further, the present invention can also be realized in the form of an image processing apparatus that detects a character string in a target image, a character string detection method, a computer program for detecting a character string, and the like as in the embodiment of FIG. By adopting such a form, it is possible to appropriately detect a character string in various images. The following configuration can be adopted as such an image processing apparatus. For example, an image processing apparatus that processes target image data representing a target image includes an area detection unit that detects a character string area representing a character string from the target image. The region detection unit has the same characteristics as the region classification unit shown in the application examples 3, 4, and 5 described above.

Modification 6:
In each of the above embodiments, a part of the configuration realized by hardware may be replaced with software, and conversely, part or all of the configuration realized by software may be replaced with hardware. Also good. For example, the function of the area classification unit 33 in FIG. 1 may be realized by a hardware circuit having a logic circuit.

  In addition, when part or all of the functions of the present invention are realized by software, the software (computer program) can be provided in a form stored in a computer-readable recording medium. In the present invention, the “computer-readable recording medium” is not limited to a portable recording medium such as a flexible disk or a CD-ROM, but an internal storage device in a computer such as various RAMs and ROMs, a hard disk, and the like. An external storage device fixed to the computer is also included.

1 is an explanatory diagram illustrating a schematic configuration of a printer 10 as an embodiment of an image processing apparatus of the present application. It is a flowchart which shows the flow of an image duplication process. It is the schematic which shows an image duplication process. It is a flowchart which shows the procedure of the area | region classification process shown to step S122 of FIG. It is the schematic of expansion and contraction of a figure. It is the schematic which shows the connection of the character pixel by expansion and contraction of a figure. It is the schematic which shows the connection of the character pixel by expansion and contraction of a figure. It is a flowchart which shows the procedure of the image compositing process shown to step S132 of FIG. It is the schematic of character string area expansion. It is the schematic of another Example of character area expansion. It is the schematic which shows another Example of a character string detection. The flowchart which shows the flow of a pixel classification process. It is the schematic which shows the object image data SI and the attention pixel. It is the schematic of object image data SI and 1st integral data ID1. It is the schematic which shows average value calculation using 1st integral data ID1. It is a flowchart which shows the flow of an area | region determination process (attribute determination process). Explanatory drawing which shows the example of the luminance value in a partial area | region. Explanatory drawing which shows the example of the histogram of dispersion | distribution. Explanatory drawing which shows an example of the partial area | region showing a halftone dot part. It is a flowchart which shows another Example of an area | region determination process (attribute determination process). Explanatory drawing which shows the example of the luminance value in three partial area | regions. Explanatory drawing which shows the example of the histogram of dispersion | distribution. It is the schematic of the determination result of the image part IP showing a character and a halftone dot.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Printer 20 ... Control unit 30 ... CPU
DESCRIPTION OF SYMBOLS 31 ... Image input part 32 ... Integration data generation part 33 ... Area classification | category part 35 ... Print control part 36 ... Image composition part 40 ... RAM
50 ... ROM
DESCRIPTION OF SYMBOLS 60 ... Carriage moving mechanism 62 ... Carriage motor 64 ... Drive belt 66 ... Sliding shaft 70 ... Carriage 71 ... Ink head 72 ... Ink cartridge 80 ... Paper feed mechanism 82 ... Roller 84 ... Motor 86 ... Platen 91 ... Scanner 96 ... Operation panel 332 ... Feature amount calculation unit 334 ... Area determination unit (attribute determination unit)
336: Character string detection unit 338: Margin detection unit P ... Printing paper ID1 ... First integration data ID2 ... Second integration data

Claims (8)

An image processing apparatus for processing target image data representing a target image,
A region classification unit for detecting a halftone dot region representing a halftone dot image and a character string region different from the halftone dot region and representing a character string from the target image;
An image processing unit for enlarging the character string area in the target image;
An image processing apparatus comprising: The image processing apparatus according to claim 1,
The area classification unit detects an area representing the halftone dot image as the halftone dot area regardless of whether or not the halftone dot image includes characters. The image processing apparatus according to claim 1 or 2,
The area classification unit detects a margin area,
The image processing unit enlarges the character string area within the margin area adjacent to the character string area.
Image processing device. An image processing apparatus according to any one of claims 1 to 3,
The region classification unit detects a character pixel representing a character from a plurality of pixels included in the target image data, and includes a region including a connected region obtained by connecting character pixels arranged apart from each other. Detect as a string area,
Image processing device. The image processing apparatus according to claim 4,
The area classification unit is an image processing device that obtains the connected area by anisotropic connection that connects the character pixels along a predetermined direction. The image processing apparatus according to claim 5,
The region classification unit is configured to determine a length along the predetermined direction of a region where the character pixels are connected, for each result of N types (N is an integer of 2 or more) of anisotropic connections in which the predetermined directions are different from each other. An image processing apparatus that determines a representative value of the character string and detects the character string region in accordance with the anisotropic connection having the longest representative value. An image processing method for processing target image data representing a target image,
Detecting, from the target image, a halftone dot region representing a halftone dot image, and a character string region different from the halftone dot region and representing a character string;
Expanding the character string region in the target image;
An image processing method. A computer program for processing target image data representing a target image,
A function for detecting, from the target image, a halftone dot region representing a halftone dot image and a character string region different from the halftone dot region and representing a character string;
A function of enlarging the character string area in the target image;
A computer program that causes a computer to realize

Images