JP5003445B2 - Image processing apparatus, character area specifying method, and character area specifying program - Google Patents

Description

  The present invention relates to an image processing device for specifying a character region from an image composed of a plurality of pixels, a character region specifying method using the image processing device, and a character region specifying program, and particularly includes complicated and diverse backgrounds. The present invention relates to an image processing apparatus that accurately generates a character line rectangle from an image, and that accurately specifies a character line rectangle, a character area specifying method using the image processing apparatus, and a character area specifying program.

  In recent years, with the progress of computerization of information, there is an increasing demand for storing or transmitting documents in electronic form instead of paper. Therefore, in an image processing apparatus such as an MFP (Multi Function Peripheral) that acquires image data, a function of directly transmitting image data obtained by scanning without attaching it to paper without attaching it to paper has become widespread. ing.

  By the way, since an image handled in an image processing apparatus such as an MFP is shifting from a monochrome image to a color image, the image data is becoming color image data. In the MFP, the size of color image data obtained by scanning a full-color original of a size of 297 mm × 210 mm, so-called A4 size, at a resolution of 300 dpi reaches about 25 MB. For this reason, there has been a problem that the color image data is too large to be attached to an email and transmitted.

  In order to solve this problem, image data (abbreviated as scan data) obtained by scanning is generally compressed and reduced in size before transmission. However, if the scan data is compressed with the same resolution for the entire image, the legibility of the characters included in the image is impaired. If compression is performed with a resolution high enough to ensure the legibility of characters in the image, there is a problem that the size of the scan data cannot be reduced sufficiently.

  In order to solve this problem, in Japanese Patent Application Laid-Open No. 2004-304469 (Patent Document 1) filed and filed by the applicant of the present application, scan data with different resolutions and different compression methods for each region in an image. A compression method such as a so-called compact PDF (Portable Document Format) has been proposed. According to this method, the compact PDF (1) executes a process for determining the area of the scan data, separates the character area and the part other than the character, and (2) the high resolution of the character area. Quantization processing, character of the same color is integrated to determine the color of the character, reversible compression such as MMR (Modified Modified-Read) compression, and (3) lowering the resolution for the part other than the character It is created by a procedure of irreversible compression such as JPEG (Joint Photographic Experts Group) compression and (4) superimposing the JPEG layer and the MMR layer. In this way, it is possible to generate a PDF file that achieves both character legibility and data compressibility.

  Since the compact PDF is generated by the above procedure, it is important to accurately extract the character region from the scanned image data in generating the compact PDF. In particular, a problem in extracting characters from an image including a complicated and diverse background is that the background region is mistakenly extracted as characters. In order to solve this problem, there is a method of performing character determination based on the likelihood of a rectangle formed by connecting pixels, taking advantage of the fact that characters often exist in line units. Yes.

  For example, in the area attribute identification method described in Japanese Patent Laid-Open No. 5-73718 (Patent Document 2), the connection state of all black pixels in a character area is examined, and a rectangular shape circumscribing a cluster where black pixels are connected. Detect coordinates. A circumscribed rectangle closest to each circumscribed rectangle is detected and its positional relationship is detected. If it is connected in the horizontal direction, one horizontal direction coupling counter is recommended. If it is connected in the vertical direction, one vertical direction combination counter is recommended. When processing is completed for all circumscribed rectangles, the values of the horizontal direction counter and the vertical direction counter are compared. If there are many horizontal direction counters, the horizontal writing area is determined, and if there are many vertical direction counters, the vertical writing area is determined. Is determined.

  In the document image region extraction method described in Japanese Patent Laid-Open No. 5-166000 (Patent Document 3), a region image is created by region image creation processing, and a label image is created by label image creation processing using the region image. create. In the document image creation process, all regions other than the character region are made white from the label image, and a character image only of characters is created. Then, in the adjacent area search process, the character area located closest to each of the upper, lower, left, and right is searched for each character area. Based on this result, the character areas are combined in the row direction or the column direction by the writing direction combination processing. Further, a plurality of character strings having the same writing direction are combined in the grouping combination process. Finally, in the area integration process, the area of the document image is extracted by obtaining the minimum rectangular area surrounding the character areas combined in the grouping and combining process.

  Furthermore, Japanese Patent Application Laid-Open No. 2007-193750 (Patent Document 4) separates a figure area mainly including a figure and a graph from a scanned image and a text area from the scanned image. An image processing apparatus that performs character determination in units of rectangles obtained by connecting and connecting black pixels is described. This image processing apparatus extracts a circumscribed rectangle of continuous black pixels by performing a labeling process for the figure region without connecting the black pixels, and performs character determination in units of the rectangles.

Thus, a row rectangle forming method is known in which margins are extracted and pixels are connected to the remaining region to generate a row rectangle. It is known that when this blank space extraction process is performed, the space between the character lines is separated by a blank space, so that the formation accuracy of the line rectangle is good. Also, a character area determination method for determining a character rectangle if the aspect ratio of the connected rectangles is equal to or greater than a prescribed value is also known.
JP 2004-304469 A JP-A-5-73718 Japanese Patent Laid-Open No. 5-166000 JP 2007-193750 A

  However, in the above conventional line rectangle forming method, in the case of a character line that does not have a character line around it or a character line that exists at the end of the text area, a mere space between characters may be erroneously determined as a blank space. The character line is divided by this erroneous determination, and an accurate character line cannot be formed even if pixels are connected. Further, in the conventional character determination method, since a vertically long and horizontally long rectangle exists in the background area, a part of the background may be extracted as a character.

  The present invention has been made to solve the above problems, and a main object of the present invention is to generate a character line rectangle more accurately from an image including a complex and diverse background, and more accurately An object is to provide an image processing device for specifying a line rectangle, a character region specifying method using the image processing device, and a character region specifying program.

According to an aspect of the present invention, there is provided an image processing device for specifying a character area from an image composed of a plurality of pixels, a generating means for generating a binary image based on the image, and a margin area from the binary image. a first extracting means for extracting and other non-margin area, by performing the expansion processing of the pixels in the first direction and a second direction perpendicular to it in a non-blank area, the first row rectangular Each of the first line rectangles in the character line direction determined by the determination means, the expansion means for generating the image line, the determination means for determining the character line direction of the binary image based on the aspect ratio of the first line rectangle First connecting means for generating a second row rectangle by connecting and first specifying means for specifying a character area based on an area corresponding to the second row rectangle in the binary image is provided.

  According to this aspect, the first connecting means generates a line rectangle even from a character line that does not have a character line around it or a character line at the end of the text area. For this reason, the character line rectangle can be formed more accurately than the conventional method, and the extraction accuracy of the character region is improved.

Preferably, the determination means includes, among the first row rectangles, the number of row rectangles satisfying a first predetermined condition indicating that the aspect ratio is a row rectangle having a horizontally long aspect ratio, and a row having a vertically long aspect ratio. Calculating means for calculating the number of row rectangles satisfying a second predetermined condition indicating that the image is a rectangle, and first determining means for determining the character line direction of the binary image based on the two numbers. .

Preferably, the first specifying means extracts at least one small rectangle included in the second row rectangle, and for each second row rectangle, the aspect ratio of the second row rectangle to the aspect ratio of the second row rectangle Second determining means for determining an area corresponding to the second row rectangle as a character area when the ratio of the number of small rectangles is within a predetermined range .

  In this case, after extracting the second row rectangle, it is further determined whether or not the row rectangle is a character region using the small rectangle information in the row rectangle. Can be specified. In other words, the extraction accuracy of the character region is improved, and background misextraction is reduced.

Preferably, the first specifying unit includes a second extraction unit that extracts at least one small rectangle included in the second row rectangle, and a small row included in the second row rectangle for each second row rectangle. And third determining means for determining an area corresponding to the second row rectangle as a character area when the variation in the size of the rectangle is equal to or less than a predetermined value .

  Preferably, the third row rectangle is obtained by connecting the first row rectangles other than the region specified as the character region by the first specifying unit in a direction perpendicular to the character line direction determined by the determination unit. The apparatus further includes second connecting means to be generated and second specifying means for specifying a character area based on an area corresponding to the third row rectangle in the binary image.

Preferably, the generation unit generates a plurality of types of binary images corresponding to each of a plurality of colors obtained by performing a color reduction process on the image, and includes a first extraction unit, an expansion unit, and a determination unit. And the first connecting means performs processing on each binary image, and the first specifying means performs character processing based on a union of areas corresponding to the second row rectangles in the respective binary images. Identify the area.

According to another aspect of the present invention, there is provided a character area specifying method using an image processing apparatus for specifying a character area from an image composed of a plurality of pixels, the image processing apparatus including a control unit, The character region specifying method includes a step in which a control unit generates a binary image based on an image, a step in which the control unit extracts a blank region and other non-margin regions from the binary image, and a control unit , by performing the expansion processing of the pixels in the first direction and a second direction perpendicular to it in a non-blank area, and generating a first row rectangle, respectively, the control unit, of the first row rectangular A step of determining the character line direction of the binary image based on the aspect ratio , a step of generating a second line rectangle by connecting the first line rectangles in the character line direction, and a control unit; Is paired with the second row rectangle in the binary image. And a step of specifying a character region based on the region.
Preferably, the determining step includes, among the first row rectangles, the number of row rectangles satisfying the first predetermined condition indicating that the aspect ratio is a row rectangle having a horizontally long aspect ratio, and the aspect ratio being vertically long. A step of calculating the number of row rectangles satisfying a second predetermined condition indicating a row rectangle, and a step of determining a character row direction of the binary image based on the two numbers.
Preferably, the step of specifying the character region includes: extracting at least one small rectangle included in the second row rectangle; and, for each second row rectangle, the small rectangle with respect to the aspect ratio of the second row rectangle. Determining a region corresponding to the second row rectangle as a character region when the number ratio is within a predetermined range.
Preferably, the step of specifying the character region includes the step of extracting at least one small rectangle included in the second row rectangle, and the size of the small rectangle included in the second row rectangle for each second row rectangle. And determining a region corresponding to the second row rectangle as a character region when the variation of is less than or equal to a predetermined value.
Preferably, in the character region specifying method, the first row rectangles other than the region specified as the character region by the step of specifying the character region are connected in a direction perpendicular to the character line direction determined in the determining step. Thus, the method further includes the step of generating the third row rectangle and the step of specifying the character region based on the region corresponding to the third row rectangle in the binary image.
Preferably, in the generating step of generating a binary image, a plurality of types of binary images corresponding to each of a plurality of colors obtained by performing a color reduction process on the image are generated, and a margin area and other areas are generated. In each of the step of extracting the non-margin area, the step of generating the first row rectangle, the step of determining, and the step of generating the second row rectangle, each binary image is processed. In the step of specifying the character region, the character region is specified based on the union of the regions corresponding to the second row rectangles in the respective binary images.

According to still another aspect of the present invention, there is provided a character area specifying program for causing a computer to specify a character area from an image composed of a plurality of pixels. The program is a binary image based on an image. Generating a margin area and other non-margin areas from the binary image, and performing pixel expansion processing in the first direction and a second direction orthogonal thereto in the non-margin area it allows generating a first row rectangle respectively, and determining the character line direction of the binary image on the basis of the aspect ratio of the first row rectangular, the first line rectangle between the character line direction The step of generating the second row rectangle by connecting the characters and the step of specifying the character region based on the region corresponding to the second row rectangle in the binary image are executed.
Preferably, the determining step includes, among the first row rectangles, the number of row rectangles satisfying the first predetermined condition indicating that the aspect ratio is a row rectangle having a horizontally long aspect ratio, and the aspect ratio being vertically long. A step of calculating the number of row rectangles satisfying a second predetermined condition indicating a row rectangle, and a step of determining a character row direction of the binary image based on the two numbers.
Preferably, the step of specifying the character region includes: extracting at least one small rectangle included in the second row rectangle; and, for each second row rectangle, the small rectangle with respect to the aspect ratio of the second row rectangle. Determining a region corresponding to the second row rectangle as a character region when the number ratio is within a predetermined range.
Preferably, the step of specifying the character region includes the step of extracting at least one small rectangle included in the second row rectangle, and the size of the small rectangle included in the second row rectangle for each second row rectangle. And determining a region corresponding to the second row rectangle as a character region when the variation of is less than or equal to a predetermined value.
Preferably, the program connects the first line rectangles other than the area specified as the character area by the step of specifying the character area in a direction perpendicular to the character line direction determined in the determination step. And a step of generating a third row rectangle and a step of specifying a character region based on a region corresponding to the third row rectangle in the binary image.
Preferably, in the generating step of generating a binary image, a plurality of types of binary images corresponding to each of a plurality of colors obtained by performing a color reduction process on the image are generated, and a margin area and other areas are generated. In each of the step of extracting the non-margin area, the step of generating the first row rectangle, the step of determining, and the step of generating the second row rectangle, each binary image is processed. In the step of specifying the character region, the character region is specified based on the union of the regions corresponding to the second row rectangles in the respective binary images.

  As described above, according to the present invention, it is possible to generate a character line rectangle more accurately from an image including a complicated and diverse background and to specify the character line rectangle more accurately.

  Embodiments of the present invention will be described below with reference to the drawings. In the following description, the same parts and components are denoted by the same reference numerals. Their names and functions are also the same.

  In the present embodiment, the image processing apparatus according to the present invention is typically an MFP (Multi Function Peripheral) integrated with a copy function, a scan function, a FAX transmission function, and the like. However, the image processing apparatus according to the present invention is not limited to an MFP, and may be any other apparatus as long as it is provided with a means for processing input image data, such as a general personal computer. Also good.

<Hardware configuration>
FIG. 1 is a diagram showing a specific example of the hardware configuration of the MFP 10 according to the present embodiment. Referring to FIG. 1, an MFP 10 according to the present embodiment processes an image (image data) composed of a plurality of pixels (pixel data), and includes a scan processing unit 1, an input image processing unit 2, and the like. A storage unit 3, a CPU (Central Processing Unit) 4, a network I / F (interface) 5, an output image processing unit 6, an engine unit 7, a modem / NCU (Network Control Unit) 8, and an operation unit 9.

  The scan processing unit 1 scans and reads a set original in accordance with a control signal from the CPU 4, and outputs image data to the input image processing unit 2. The input image processing unit 2 calculates a value such as RGB data for each pixel of the image data input from the scan processing unit 1 according to the control signal, and outputs the value to the CPU 4 or stores it in the storage unit 3.

  The storage unit 3 includes an electronic memory such as DRAM (Dynamic Random Access Memory) or SRAM (Static Random Access Memory), and a magnetic memory such as a hard disk, and holds programs and image data. The storage unit 3 is also used as a work area when the CPU 4 executes a program.

  The CPU 4 executes a program stored in the storage unit 3. Based on the operation signal input from the operation unit 9, the CPU 4 outputs necessary control signals to each unit to control the entire MFP 10. For example, the CPU 4 detects an operation key, displays an operation panel, changes input data to an image file, creates an e-mail, and the like. The CPU 4 executes a program stored in the storage unit 3, performs image processing on the image data held in the storage unit 3, and stores the storage unit 3, the network I / F unit 5, or the modem / NCU 8. A control signal or the like is output. The CPU 4 executes processes such as color conversion, color correction, resolution conversion, and area specification according to the present embodiment on the image data input from the input image processing unit 2. The processed data is held in the storage unit 3.

  The network I / F unit 5 is an I / F for transmitting an electronic mail or the like to another device via the network, and creates a data packet according to a protocol. The network I / F unit 5 transmits the image data input from the CPU 4 or the image data read from the storage unit 3 to another device via the network according to the control signal.

  The output image processing unit 6 reads the image data held in the storage unit 3 in accordance with the control signal, performs screen control, smoothing processing, etc. on the image, and outputs the processed image data to the engine unit 7. To do.

  The engine unit 7 generates a toner image based on the image data input from the output image processing unit 6 according to the control signal, and prints the image by transferring the toner image onto the set printing paper. When the MFP 10 is a color MFP that outputs a color image, the engine unit 7 generates a toner image using toners of four colors, yellow, magenta, cyan, and black.

  The modem / NCU 8 controls communication via a telephone line according to modulation / demodulation for facsimile transmission / reception, a facsimile communication protocol, and the like. The modem / NCU 8 transmits the image data input from the CPU 4 or the image data read from the storage unit 3 to another device via a telephone line in accordance with the control signal.

  The operation unit 9 includes an operation key and a display unit, and functions as a user I / F to input a destination from the user, select a scan condition, select an image file format, start / interrupt processing, and the like. Accept the operation. The operation unit 9 outputs an operation signal based on a user operation to the CPU 4.

<Functional configuration>
FIG. 2 is a block diagram showing a specific example of a functional configuration for compressing image data and creating a PDF (Portable Document Format) file in the MFP 10 according to the present embodiment. Each unit shown in FIG. 2 is a function realized mainly by the CPU 4 executing a program stored in the storage unit 3, but some functions are, for example, the scan processing unit 1, the input image processing unit 2, and the like. It may be realized by other dedicated hardware configuration.

  Referring to FIG. 2, functions for creating a PDF file in MFP 10 according to the present embodiment include an image data acquisition unit 101, a preprocessing unit 103, a photo determination unit 104, a character region specification unit 105, and the like. The reversible compression unit 107, the resolution reduction unit 109, the irreversible compression unit 111, and the PDF conversion unit 113 are included.

The image data acquisition unit 101 acquires the image data generated by the scan processing unit 1 and performs TIFF (Tagged Image File Format), JPEG (Joint Photographic Experts).
Group), BMP (Bit MaP), and other data formats are input to the preprocessing unit 103. In other words, the image data acquisition unit 101 is a scanner portion of the MFP 10 and is a portion that reads a document and outputs image data.

  The pre-processing unit 103 performs processing such as image format conversion, resolution conversion, and background removal on the image data input from the image data acquisition unit 101 as pre-processing for specifying a character region. Input to the unit 105. The preprocessing unit 103 performs HSL (hue, saturation, brightness) conversion, binarization of lightness, and labeling. In labeling, a rectangle number, the upper left coordinates of the circumscribed rectangle, and the lower right coordinates are given to the connected pixels obtained by the binarization of the lightness to form a rectangle.

  The photo determination unit 104 determines a photo area. The reason for discriminating the photographic area is to change the character extraction processing method between the photographic area and an area other than the photographic area. When attempting to extract characters from a photo area, it is often mistakenly determined that a non-character is a character, and it is necessary to more accurately determine characters in the photo area. However, the determination method of the photograph area is not limited here. Typically, among the rectangles obtained by the preprocessing unit 103, a rectangle having a predetermined size that is determined to have a possibility of being photographed is used if the number of colors is large by using hue data. There is a method of determining that it is.

  The character area specifying unit 105 performs a character area specifying process on image data of an area other than the photo area input from the photo determining unit 104. Character area specifying processing includes color reduction processing (binarization processing), margin extraction processing, expansion processing, character line direction determination processing, first connection processing, first character line determination processing, second connection processing, character Includes color determination processing. The character area specifying unit 105 executes such a character area specifying process to extract a character (character and ruled line) area in the image. Then, the character area specifying unit 105 separates the character area from a background area such as a photograph, a figure, or a graph other than the character area. The character region here includes a character region existing in the figure such as a photograph, a figure, or a graph. In other words, the character area specifying unit 105 extracts a character area by performing different character area specifying processes for areas other than the photo area and the photo area, and is a block having a function for calculating a character color. is there. The character area specifying process on the area other than the photograph area will be described later.

  The image data constituting the separated background region, that is, the region not specified as the character region, is input to the lossy compression unit 111 via the resolution reduction unit 109. On the other hand, image data constituting an area identified as a character area is directly input to the lossless compression unit 107 without passing through the resolution reduction unit 109.

  The lossless compression unit 107 performs reversible compression such as an MMR (Modified Modified-Read) compression method on the image data constituting the character region input from the region specifying unit 105. In addition, the irreversible compression unit 111 performs irreversible compression such as a JPEG compression method on the image data constituting the background area whose resolution has been reduced by the resolution reduction unit 109. The image data constituting the character area and the image data constituting the background area compressed by the lossless compression unit 107 and the lossy compression unit 111 are input to the PDF conversion unit 113, and a PDF file is created based on these.

  The functional configuration of the MFP 10 shown in FIG. 2 is a configuration in the case of irreversibly compressing the image data constituting the background area by reducing the resolution of the image data when creating the PDF file. The image data to be configured may be irreversibly compressed without reducing the resolution. In that case, the function of the MFP 10 may not include the resolution reduction unit 109.

  FIG. 3 is a functional block diagram showing a functional configuration of the character area specifying unit 105. As shown in FIG. 3, the character area specifying unit 105 includes a generating unit 11, a first extracting unit 12, an expanding unit 13, a determining unit 14, a first connecting unit 17, and a first specifying unit. 18, a second connecting portion 22, and a second specifying portion 23. The determination unit 14 includes a calculation unit 15 and a first determination unit 16. The first specifying unit 18 includes a second extraction unit 19 and a second determination unit 20. The first specifying unit 18 may include a third determining unit 21 instead of the second determining unit 20 or together with the second determining unit 20. Here, each functional block constituting the character area specifying unit 105 is realized by the CPU 4 executing a program read from the storage unit 3.

  The generation unit 11 generates binary image data based on the input image data. The input image data may be image data obtained by scanning by the MFP 10 or image data input from an external device to the MFP 10. The generation unit 11 according to the present embodiment generates a plurality of types of binary image data from image data.

  More specifically, the generation unit 11 reduces the color of the image data and creates a binary image for each of a plurality of colors. The generation unit 11 generates binary image data for each pixel data based on the luminance value corresponding to the pixel data. The generation unit 11 configures the image data based on a preset threshold value of luminance value or a threshold value of luminance value determined by inspecting the target image data once. “0” (white or black) or “1” (black or white) is associated with each pixel data to be stored in the storage unit 3.

  When a threshold value is positioned between the luminance value of the character color and the luminance value of the background color in any binary image by generating a plurality of types of binary images, the pixel of the character in the binary image Data and background pixel data can be separated into “0” and “1”. Since characters in the same character line often have the same color, they are often reduced to the same color (white or black). For example, in MFP 10 according to the present embodiment, the number of color reductions is four. The method for calculating the threshold value for color reduction is not particularly limited, but there are a frequency method, a median value division method, and the like.

  The first extraction unit 12 extracts a margin area and other non-margin areas from the binary image. Here, the “margin area” refers to a blank portion where there are not more than a specified number of black pixels (the number of white pixels continues for a specified number). The “non-margin area” refers to an area other than the margin area in the image data.

  More specifically, the first extraction unit 12 scans each binary image created by the generation unit 11 in the main scanning direction and the sub-scanning direction, and there are no more than a specified number of black pixels ( A blank part (blank area) in which white pixels continue for a predetermined number or more) is extracted and filled with blank white (for example, “0”). On the other hand, the pixel data corresponding to the black pixel is set to “1” for the non-blank area where the white pixels do not continue the predetermined number or more. The specified number is, for example, the number of pixels corresponding to a length of 4.5 cm (about 600 pixels for 300 dpi). This extraction process can prevent the later-described expansion unit 13 from connecting characters of different character lines. However, as a result of performing the extraction process, in the case of a character line in which there are no black pixels (pixels associated with “1”) around the character line, the character line may be divided. A specific image will be described later.

  FIG. 4 is an image diagram showing pixel expansion in the main scanning direction. The “margin area” is an area composed of pixels associated with “−1” in FIG. 4. The “non-margin area” refers to an area composed of pixels associated with “0” or “1” in FIG.

  The pixel expansion will be described with reference to FIG. The expansion unit 13 performs at least expansion processing of black pixels (pixels whose pixel data is “1”) in the non-blank area based on the image data 20M passed from the first extraction unit 12. One first row rectangle is generated. More specifically, black pixels are expanded and connected to each other in areas that are not determined to be blank areas by the first extraction unit 12. The expansion process of the black pixel here is the peripheral after the CPU 4 expands the pixel in the horizontal direction by confirming the neighboring pixel in the main scanning direction based on the input peripheral pixel map (image data) 20. Pixel maps (image data) 20L and 20R are created. Here, the image data 20L refers to image data that has been subjected to expansion processing in the left direction. The image data 20R refers to image data that has been expanded rightward.

  Specifically, the CPU 4 performs the following processing while moving in the main scanning direction on each line in the main scanning direction. (1) The CPU 4 inspects the value of the target pixel. When the target pixel is filled as a blank pixel, the target pixel in the peripheral pixel map after being expanded in the left-right direction is set to “−1”. (2) When the pixel is not a blank pixel (white pixel and black pixel), for example, it is checked whether there is a black pixel within half the maximum character width from the target pixel to the left side (when expanding the pixel to the right) To do. Here, the maximum character width is, for example, 190 pixels.

  If there is a black pixel, the value of the pixel of interest in the peripheral pixel map after expansion to the right is set to “1”. If there is no black image pixel, 0 is set. However, if a blank pixel is found in the middle, no further search is performed. In this way, the expansion unit 13 searches the left side for each pixel of interest, that is, performs processing for expanding the black pixel in the right direction to generate the image data 21R. The expansion unit 13 performs the same process on the right side to generate the image data 21L. In this way, the peripheral pixel map is completed for all the pixels.

  FIG. 5 is an image diagram showing pixel expansion in the sub-scanning direction of the image data 21R expanded in the right direction shown in FIG. As illustrated in FIG. 5, the expansion unit 13 connects the peripheral pixel maps in the sub-scanning direction with respect to the image data 21 </ b> R, and generates image data 22 </ b> R. More specifically, the expansion unit 13 (1) scans each X coordinate in the Y direction, and detects “1” pixels existing within a first predetermined interval (one pixel in the present embodiment). Link. That is, the expansion unit 13 fills the pixel data “0” between the pixel data “1” and the pixel data “1” with the pixel data “1”. As in the expansion in the main scanning direction, the pixel data “−1” is not changed.

  FIG. 6 is an image diagram showing pixel expansion in the sub-scanning direction of the image data 21L after expansion in the left direction shown in FIG. As illustrated in FIG. 6, after the expansion unit 13 expands pixels in the sub-scanning direction for the image data 21 </ b> R, the expansion unit 13 similarly expands the pixels in the sub-scanning direction for the image data 21 </ b> L.

  FIG. 7 is an image diagram showing a union of black pixels of the image data 22L and 22R after expansion in the left-right direction shown in FIG. 5 and FIG. After the pixel expansion in the sub-scanning direction of the image data 21R / 21L, the expansion unit 13 performs (2) the left expanded image data 22L and the right pixel expanded pixel data 22R corresponding to the target pixel. It is checked whether or not the value of the peripheral pixel map of any one of the image data 22L and 22R is “1”. That is, as illustrated in FIG. 7, when the expansion unit 13 superimposes the image data 22L and 22R and any of the left and right pixel data 22L and 22R is “1”, the pixel in the horizontal direction ( It is determined that it is part of a character line in the main scanning direction) and is painted black.

  Thereafter, the expansion unit 13 rotates the image data 20M by 90 degrees, performs pixel expansion in the main scanning direction and pixel expansion in the sub-scanning direction, and concatenates character lines in the vertical direction. Let

  Specifically, as illustrated in FIG. 8, the expansion unit 13 performs a search on the left side of each pixel based on the image data 30 rotated by 90 degrees, that is, a process of expanding the black pixel in the right direction. To generate image data 31R. Then, the expansion unit 13 performs the same process on the right side of each pixel, and generates image data 31L in which the pixel is expanded leftward.

  Then, as shown in FIG. 9, the expansion unit 13 connects pixels in the sub-scanning direction with respect to the image data 31R expanded rightward to generate image data 32R. As shown in FIG. 10, the expansion unit 13 similarly performs pixel expansion in the sub-scanning direction on the image data 31L to generate image data 32L. Then, as illustrated in FIG. 11, the expansion unit 13 uses any one of the image data for the pixel data 32R after pixel expansion in the right direction and the pixel data 32L after pixel expansion in the left direction corresponding to the target pixel. It is checked whether the value of the pixel data of 32L and 32R is “1”. Then, as shown in FIG. 11, when either one of the left and right image data 32L and 32R is “1”, the expansion unit 13 determines that the pixel is a part of the horizontal character line. Fill in black.

  Returning to FIG. 3, the determination unit 14 determines the character line direction of the binary image based on the shape of the first line rectangle. Specifically, the calculation unit 15 includes the number of row rectangles whose aspect ratio satisfies the first predetermined condition and the number of row rectangles whose aspect ratio satisfies the second predetermined condition among the first row rectangles. And are calculated. Then, the first determination unit 16 determines the character line direction of the binary image based on the two types of numbers. More specifically, the determination unit 14 can reconnect the black pixels in either the horizontal or vertical direction in order to generate a more accurate character line rectangle based on a character line that is not completely formed. Judge what to do.

  Specifically, first, the calculation unit 15 searches for a rectangle having an aspect ratio (aspect ratio) of a first predetermined value (for example, 3) or more in the first row rectangle formed by the expansion unit 13. To do. Then, the calculation unit 15 searches for a rectangle having an aspect ratio (aspect ratio) equal to or less than a second predetermined value (for example, 1/3) in the first row rectangle. Next, a rectangle with an aspect ratio equal to or greater than the specified value is regarded as a character line, and it is determined whether there are more character line rectangles, a long character line rectangle or a long character line rectangle, as a whole image. The direction of the character line as the whole image is determined.

  The first connecting unit 17 generates a second line rectangle by connecting the first line rectangles in the character line direction determined by the determining unit 14. More specifically, the first connecting unit 17 connects character line rectangles (black pixels) formed by expanding the pixels in the expansion unit 13 in the character line direction determined by the determination unit 14. Connect again. Accordingly, the first connecting unit 17 generates an accurate character line rectangle by connecting the characters in the character line that have been determined as the blank area by the first extracting unit 12. At this time, it is assumed that black pixels may be connected across a blank area (pixel data of “−1”). By this processing, it becomes possible to generate a line rectangle for all the character lines.

  FIG. 12 is an image diagram showing rotation of image data. Here, when the character line direction is the horizontal direction, the following processing is executed. (1) First connecting unit 17 first rotates image data 40 by 90 degrees to obtain image data 50. (2) The first connecting unit 17 inspects the value of the target pixel.

  FIG. 13 is an image diagram showing expansion processing for the image data shown in FIG. As shown in FIG. 13, when the pixel is filled as a blank pixel, the left and right peripheral pixel maps are set to “−1”. If it is not a blank pixel (when white pixel is “0” or black pixel “1”), it is checked whether there is a black pixel within half the maximum character width (for example, 190 pixels) from the target pixel to the left side. . If there is a black pixel, the value of the left peripheral pixel map is set to “1”. When there is no black pixel, “0” is set. However, if a blank pixel is found in the middle, no further search is performed. Thus, the 1st connection part 17 acquires the image data 50R after the expansion | swelling to the right direction. As shown in FIG. 13, the first connecting unit 17 performs the same processing as described above for each pixel on the right side, that is, expands the black pixel leftward to obtain the image data 50L. The peripheral pixel map is completed for all pixels.

  FIG. 14 is an image diagram showing a first connection process between the first character line rectangles (black pixels) shown in FIG. Following the processing of (2), as shown in FIG. 14, (3) the first connecting portion 17 scans in the Y direction for each X coordinate, and the second predetermined interval (3 in the present embodiment). “1” existing within (pixel) is connected. Here, the second predetermined interval is specified to be wider than the first predetermined interval in the expansion process. Further, unlike the connection in the expansion process, even the blank pixel “−1” is filled with 1.

  Finally, (4) the first connecting unit 17 checks whether the value of the peripheral pixel map is “1” for the image data 51L / 51R expanded in the left-right direction corresponding to the target pixel, and expanded in the left-right direction. When any one of the image data 51L and 51R is “1”, it is determined that the pixel is a part of the horizontal row, and the pixel is painted black. That is, the union of the first character line rectangles (black pixels) is calculated by superimposing the image data 51L and 51R. As a result, as shown in FIG. 14, the blank space that exists between the characters in the character line is filled with black pixels, and a more accurate character line rectangle is formed.

  FIG. 15 is an image diagram showing pixel data 61 obtained by superimposing pixel data 50L and 50R obtained by performing only expansion processing on the image data shown in FIG. For reference, as shown in FIG. 15, the pixel data 61 obtained by superimposing the pixel data 50L and 50R obtained by performing only the dilation processing is insufficiently connected with black pixels, so that accurate characters are displayed. You can see that no rows have been generated. That is, in FIG. 14, the blank area between characters is filled with “1”, but in FIG. 15, the blank area between characters remains “−1” and an accurate character line rectangle is not formed. Recognize.

  On the other hand, when the character line direction is vertical, the above processing is performed without rotating the image.

  The expansion unit 13 specifies the position coordinates, the size, and the shape of the first row rectangle by acquiring the coordinates of the four corners of the rectangular region (first row rectangle) including the black pixel after performing the expansion process. . Similarly, the first connecting unit 17 obtains the coordinates of the four corners of the rectangular area (second row rectangle) including the black pixel after the connection processing, thereby obtaining the position coordinates of the second row rectangle, Identify size and shape.

  Returning to FIG. 3, the specifying unit 18 specifies a character area based on an area corresponding to the second row rectangle in the binary image. The second extraction unit 19 extracts at least one small rectangle included in the second row rectangle. When the relationship between the aspect ratio of the second row rectangle and the number of small rectangles satisfies the third predetermined condition for each second row rectangle, the second determination unit 20 satisfies the second predetermined value in the binary image. The area corresponding to the line rectangle is determined as the character area. The third determination unit 21 determines the second row in the binary image when the relationship between the size of the second row rectangle and the size of the small rectangle satisfies the fourth predetermined condition for each second row rectangle. The area corresponding to the rectangle is determined as the character area.

  In more detail, first, the second extraction unit 19 of the specifying unit 18 starts from the second row rectangle formed by the first connecting unit 17 based on the characteristic shape of the row rectangle, for example, the aspect ratio. Extract line candidate rectangles. That is, the second extraction unit 19 extracts a rectangle that is long in the character line direction and has an aspect ratio equal to or greater than a specified value (for example, 3) as a character line candidate rectangle. Next, the second determination unit 20 of the specifying unit 18 examines the number of small rectangles included in the extracted character line candidate rectangle. Then, the second determination unit 20 determines the character line candidate rectangle based on the aspect ratio of the character line candidate rectangle and whether the set of the number of small rectangles included in the character line candidate rectangle is a character line-like value. Characters are determined by classifying them into rectangles and partial rectangles of the background. Next, the third determining unit 21 of the specifying unit 18 also checks the number of small rectangles and the size of the small rectangles included in the extracted character line candidate rectangle. Then, the third determination unit 21 determines whether the combination of the aspect ratio of the character line candidate rectangle and the number of small rectangles included in the character line candidate rectangle is a character line value, or the size of the small rectangle is the same. Character line candidate rectangles are classified into character line rectangles and partial rectangles in the background depending on the size, and character determination is performed.

  FIG. 16 is an image diagram illustrating a method of determining whether or not the character area is in the specifying unit 18. As shown in FIG. 16A, the relationship between the aspect ratio (X / Y) of the second row rectangle and the number of small rectangles included in the second row rectangle satisfies the third predetermined condition. When the relationship between the aspect ratio (X / Y) of the second row rectangle and the size of the small rectangle included in the second row rectangle satisfies the fourth predetermined condition, the specifying unit 18 Determines that the second row rectangle is a character area.

  On the other hand, as shown in FIG. 16B, the relationship between the size of the second row rectangle and the size of the small rectangle included in the second row rectangle, for example, the height (vertical width) of the second row rectangle If the relationship with the length (horizontal width) of the small rectangle does not satisfy the fourth predetermined condition, the specifying unit 18 determines that the second row rectangle is not a character area. Specifically, when the ratio of the length of the small rectangle to the height of the second row rectangle is 1 or more, it is determined that the character area is not a character area. Specifically, the second row rectangle is not a character area when the size of the small rectangles included in the second row rectangle varies greatly, that is, when the vertical or horizontal standard deviation is greater than or equal to a predetermined value. Judge.

  Further, as shown in FIG. 16C, the relationship between the aspect ratio (X / Y) of the second row rectangle and the number of small rectangles included in the second row rectangle satisfies the third predetermined condition. For example, when the aspect ratio is 10, the number of small rectangles is less than 10 (the ratio of the number of small rectangles to the aspect ratio is less than 1), or the number of small rectangles is 50 or more. If it is (the ratio of the small rectangle to the aspect ratio is 5 or more), the specifying unit 18 determines that the second row rectangle is not a character area.

  Based on the above determination, a line rectangle that is likely to be a character line is identified from the second line rectangles. That is, the specification of the character line rectangle becomes more accurate, and the erroneous extraction of the character line is reduced.

  The second connecting unit 22 connects the first row rectangles other than the region specified as the character region by the specifying unit 18 in a direction perpendicular to the character line direction determined by the determining unit 14. 3 row rectangles are generated. That is, the 2nd connection part 22 performs the connection to a character line and a perpendicular direction. Here, with respect to an image obtained by omitting a line rectangular area determined as a character area by the specifying unit 18 from the binary image, the determination unit 14 determines for each first rectangle having the size of the specified threshold range. It is determined whether there is a rectangle of the same size (character width) in the direction perpendicular to the character line direction, and by sequentially connecting the rectangles of the same size, a line rectangle is generated again in the new character direction To do. By including this processing, it is possible to extract a character region more accurately even in an image in which character lines are included in both vertical and horizontal directions.

  And the 2nd specific | specification part 23 specifies a character area | region based on the area | region corresponding to the 3rd row rectangle in the said binary image. And each said part processes each produced | generated binary image, and the 1st specific | specification part 18 and the 2nd specific | specification part 23 are set to the said 2nd row rectangle in each said binary image. A character area is specified based on the union of the corresponding areas.

  Here, the character color determination unit determines the color of the portion determined to be a character. In the determination method, the RGB values in each character line are averaged.

<PDF processing>
FIG. 17 is a flowchart showing a processing procedure for creating a PDF file by compressing image data in the MFP 10 according to the present embodiment. The process shown in the flowchart of FIG. 17 is a process realized mainly by the CPU 4 executing a program stored in the storage unit 3 to control each unit shown in FIGS. 2 and 3.

  That is, referring to FIG. 17, in MFP 10 according to the present embodiment, first, image data acquisition unit 101 acquires image data (step 100; step is hereinafter abbreviated as S). Then, after the preprocessing in the preprocessing unit 103 is performed on the acquired image data, region specifying processing is performed in the region specifying unit 105 (S300). The image data is subjected to a compression process suitable for each area according to the determination result, and the PDF conversion process is executed in the PDF conversion unit 113 (S500), so that the image data is compressed and a PDF file is generated. Created.

  That is, in S500, reversible compression processing such as the MMR compression method is performed in the reversible compression unit 107 without reducing the resolution for the image data constituting the region determined as the character region in S300. In addition, the image data constituting the area determined as the background area in S300 is subjected to resolution conversion so as to reduce the resolution in the resolution reduction unit 109 in S500, and then the JPEG compression method is used in the lossy compression unit 111. Irreversible compression processing is performed. In S500, the irreversible compression process may be performed without reducing the resolution on the image data constituting the area determined as the background area.

  The PDF conversion process in S500 employs a general process for creating a so-called compact PDF file as described above, and is not a process limited in the present invention. Hereinafter, the processing procedure of the area specifying process (S300), which is a feature of the present invention, will be described in detail.

<Character area identification processing>
FIG. 18 is a flowchart showing the processing procedure of the character area specifying process S300. Referring to FIG. 18, the CPU 4 or the input image processing unit 2 first performs a binary image generation process on the input image data (S310). More specifically, the CPU 4 or the input image processing unit 2 performs a color reduction process on the input image data, and then performs a binarization process based on a plurality of types of threshold values (S310). In the binary image generation process here, the CPU 4 or the input image processing unit 2 generates a binary image for each of a plurality of types of colors for one image data. In MFP 10 according to the present embodiment, based on the color image shown in FIG. 19 stored in storage unit 3, CPU 4 or input image processing unit 2 uses FIG. ) To generate four types of binary images shown in FIG. Then, the process proceeds to a blank area extraction process S320.

  FIG. 21 is a flowchart showing the processing procedure of the blank area extraction processing S320. As shown in FIG. 21, the CPU 4 counts the number of pixels of a specific color (here, white pixels) as pixels that do not constitute a continuous character in the main scanning direction (x direction) for the target area. When a number of consecutive white pixels is detected, the continuous white pixels are painted with a specific color (S321). In addition, although the pixel which does not comprise the character is specifically assumed to be a pixel of a specific color that is a white pixel, the color of the target pixel may not be limited. It may be a pixel of a color other than the constituent pixel color.

  Similarly, in the sub-scanning direction, white pixels continuous for a predetermined number or more are detected, and the number of consecutive white pixels in the main scanning direction is checked with respect to the continuous white pixels (S322). The prescribed number is also determined in the same manner. Specifically, the prescribed number is 2 dots in the main scanning direction, 150 dots or more in the sub-scanning direction, and the like. Then, returning to FIG. 18, the process proceeds to the expansion process S330.

  FIG. 22 is a flowchart showing the processing procedure of the expansion processing S330. As shown in FIG. 22, in the expansion process S330, pixels of a specific color constituting the character (here, black pixels) are expanded and neighboring pixels are connected. Here, the pixels constituting the character are assumed to be pixels of a specific color that is specifically a black pixel, but the color of the target pixel may not be limited. For example, it constitutes the background. It may be a pixel of a color other than the pixel color.

  With reference to FIG. 22, first, as neighboring pixels, specifically, adjacent black pixels within a predetermined distance or less in the main scanning direction of the target region are detected (S331). More specifically, the image is scanned in the main scanning direction (x direction), and a black pixel is searched for a range of ½ of the maximum left and right character width (for example, 190 pixels) of the x coordinate position for a certain x coordinate. The array value at the y coordinate where the pixel is detected is set to 1, and the search for black pixels is repeated in the y direction sequentially from the y coordinate to the coordinate value obtained by subtracting 1 from the image height.

  However, if a line spacing is detected midway, no further search in the y direction is performed. Furthermore, the array generated with respect to the certain x coordinate is scanned, and if the sequence of array values 0 is equal to or less than a specified number, the array value of the sequences is rewritten to 1. In this embodiment, as a method for detecting neighboring pixels, a method for detecting whether or not a pixel interval is equal to or smaller than a predetermined distance is shown. However, neighboring pixels are detected by other methods. May be.

  Next, black pixels are expanded in the sub-scanning direction (y direction) (S332). More specifically, the image is scanned in the y direction with respect to the x coordinate scanned in the main scanning direction (x direction), and a black pixel is searched. Then, if the array value of the array generated for the x coordinate is 1 in the range of ½ of the maximum vertical character width (for example, 190 pixels) of the detected black pixel, the white pixels in the range are black. It is filled with.

  Next, the target region is rotated 90 degrees (S333), and similarly, adjacent black pixels within a predetermined distance in the main scanning direction are detected (S334), and these black pixels are expanded in the sub-scanning direction. (S335). FIG. 23 is an image diagram showing image data in a state in which a union of first row rectangles is acquired by superimposing an image after expansion processing before rotation and an image after expansion processing after rotation. Thereafter, returning to FIG. 18, the character line direction determination processing S340 is executed.

  FIG. 24 is a flowchart showing the processing procedure of the character line direction determination processing S340. As shown in FIG. 24, the CPU 4 performs labeling in order to obtain the minimum rectangular area surrounding the black pixel group connected by the above-described processing, and the coordinate value of the minimum rectangle surrounding the connected character obtained by the labeling. Is acquired (S341). Note that a general method is adopted as the labeling method here.

  Next, for each rectangular region obtained by the labeling, the length of the short side and the ratio of the length of the short side to the long side are calculated, and only the rectangle satisfying the predetermined condition is extracted (S342). The number of vertically and horizontally long rectangles is counted as the entire image (S343). Then, it is determined whether there are more vertically long rectangles or horizontally long rectangles (S344). If there are more vertically long rectangles (YES in S344), the character line is determined in the vertical direction (S345). If there are not more vertically long rectangles (NO in S344), the character line is determined in the horizontal direction (S346). Thereafter, returning to FIG. 18, the first connection process S350 is executed.

  FIG. 25 is a flowchart showing the processing procedure of the first connection processing S350. As shown in FIG. 25, the CPU 4 generates a second line rectangle by connecting the first line rectangles in the character line direction determined in the determination process. More specifically, the CPU 4 first expands the black pixels in the left-right direction to acquire image data (S351). Then, the left and right image data are overlapped to calculate the union of black pixels, and then “1” existing within the second predetermined interval in the character direction is connected (S352). Here, the second predetermined interval is specified to be wider than the first predetermined interval in the expansion process. Further, unlike the connection in the expansion process, even if the space between characters is a blank pixel “−1”, it is filled with the character pixel “1”. Thereafter, returning to FIG. 18, the first character determination process S360 is executed.

  FIG. 26 is a flowchart showing the processing procedure of the first character determination processing S360. As shown in FIG. 26, for each rectangular area, the CPU 4 calculates the length of the short side and the ratio of the length of the short side to the long side, and extracts only the second row rectangle that satisfies the predetermined condition. (S361). The CPU 4 extracts at least one small rectangle included in the second row rectangle (S362). For each second row rectangle, the CPU 4 sets the region corresponding to the second row rectangle as the character region when the relationship between the aspect ratio of the second row rectangle and the number of small rectangles satisfies the third predetermined condition. (S363). In addition, for each second row rectangle, when the relationship between the size of the second row rectangle and the size of the small rectangle satisfies the fourth predetermined condition, the CPU 4 selects an area corresponding to the second row rectangle. It may be determined as a character area.

  FIG. 27 is an image diagram showing an image after the first connection processing in the character line direction. As shown in FIGS. 26 and 27, compared to the first row rectangle of the image after expansion, the second row rectangle of the image after concatenation has no portion where a blank space is mixed in the character line. . That is, a more accurate character line area is acquired. Thereafter, returning to FIG. 18, the second connection process S370 is executed.

  FIG. 28 is a flowchart illustrating a processing procedure of the second connection processing S370. As shown in FIG. 28, the CPU 4 performs labeling in an area other than the area specified as the character area by the specifying unit 18 in order to obtain a rectangular area surrounding the black pixel group connected by the above-described processing. The coordinate value of the minimum rectangle surrounding the connected characters obtained by the above is acquired (S371). Note that a general method is adopted as the labeling method here. Next, for each rectangular region obtained by the labeling, the length of the short side and the ratio of the length of the short side to the long side are calculated, and only the rectangle satisfying the predetermined condition is extracted (S372). Here, the CPU 4 may generate the third line rectangle by connecting the rectangles satisfying the predetermined condition in a direction perpendicular to the determined character line direction. And CPU4 specifies a character area based on the area | region corresponding to the 3rd row rectangle in a binary image similarly to 1st character determination process S360.

  Thereafter, returning to FIG. 18, it is determined whether or not the processing has been completed for all the binary images (S380). If processing has not been completed for all binary images (NO in S380), the processing from S320 to S370 is repeated for the next binary image. On the other hand, when the processing is completed for all the binary images (YES in S380), the character color determination processing S390 is executed.

  FIG. 29 is a flowchart illustrating a processing procedure for character color determination processing. As shown in FIG. 29, the color of the character is determined by referring to the color data (RGB data) of the original image corresponding to the character area. The CPU 4 calculates the average value of RGB data for each second row rectangle determined as a character area (S391), and assigns one color to one character area (S392). If the interval between two adjacent rectangular regions determined to be character regions is equal to or smaller than a predetermined interval and the color difference assigned to the rectangular regions is equal to or smaller than a predetermined value, these rectangular regions are integrated. Also good. By doing so, the number of rectangular areas that are character areas can be reduced, and the size of the created PDF data can be reduced. In addition, the speed of the creation process can be increased.

  As described above, the processing for all the binary images is completed, and the area specifying process S300 in creating compact PDF data is completed. FIG. 30 is an image diagram showing an image of a final character area before the creation of compact PDF data.

  As described above, in the area specifying process according to the present embodiment, the character determination process is branched for the text area and the figure area included in the document, and the conventional black pixel is connected to the text area. While character determination is performed in units, character determination is performed for each rectangular area obtained by labeling without connecting black pixels in the figure area.

  In this way, by not connecting black pixels in the figure area, lines and dots that are considered to exist in the vicinity of the characters in the figure area are connected to the pixels that make up the character, thereby reducing the accuracy of character determination. Is prevented. As a result, the character area in the figure area is determined with high accuracy. Moreover, the processing speed is also ensured by dividing the figure area and the text area and performing the conventional character determination for the text area.

<Summary>
The features of the image processing apparatus (MFP 10) according to the present embodiment will be summarized below. The method for forming a character line rectangle according to the present embodiment uses information on a rectangular group (vertically long rectangle or horizontally long rectangle) formed once in a row and determined that there is a space between characters. The character line direction is determined as to whether the character line that has been subjected to the maximum connection in the vertical direction or the horizontal direction is determined, and the connection process is performed again, thereby accurately generating a line rectangle. In addition, in order to accurately extract the character line in the direction perpendicular to the character line direction when the entire image is viewed, scanning is performed in the direction perpendicular to the character line direction when viewed in the entire image, and the specified threshold value is exceeded. It is determined whether there is a rectangle having the same size in the vertical direction. If there are rectangles having the same size, row rectangles are generated by sequentially connecting the rectangles.

  Then, as a method of determining the character according to the present embodiment, after extracting a rectangle that seems to be a character line having an aspect ratio of a specified value or more from the formed rectangle, a small rectangle (character unit) included in the line rectangle The line rectangle having a variation in the size of the image is determined to be a part of the background, thereby accurately determining the character. In addition, regarding the combination of the number of small rectangles (corresponding to character units) contained in the row rectangle and the aspect ratio of the row rectangle, the rectangle that has too few or too many small rectangles relative to the row rectangle aspect ratio is By determining that it is a part of the background, the character is accurately determined.

  Specifically, the MFP 10 according to the present embodiment for achieving the above object has a configuration for realizing the following functions. (1) The color of the scanned image is reduced, and a binary image is created for each color (for the purpose of separating the character and the background around the character). (2) For each created image, the image is scanned in the main scanning direction and the sub-scanning direction, and a blank portion having no black pixels is extracted. (3) The pixels are expanded and connected to each remaining area. (4) Among connected rectangles, rectangles with an aspect ratio equal to or greater than a specified value (for example, 3) are regarded as rectangles in a line lump. Determine. (5) With respect to the obtained direction (character line direction), the pixels are expanded again and connected, and a character line is formed from a character that cannot be formed because a blank space is extracted between characters. At this time, the region extracted as the blank in (2) may be straddled. (6) A rectangle having an aspect ratio of a specified value (for example, 3) or more is extracted as a character line candidate rectangle from the connected rectangles. (7) The number of small rectangles and the size of the small rectangles included in the extracted character line candidate rectangles are examined. (8) A character line depending on the aspect ratio of the character line candidate rectangle and whether the combination of the number of small rectangles included in the character line candidate rectangle is a character line-like value or the size of the small rectangles is the same size. The candidate rectangle is classified into a character line rectangle and a partial rectangle of the background, and character determination is performed. (9) In the direction perpendicular to the character line direction, each image obtained by omitting the rectangle determined to be a character from the image formed in (2) is scanned, and the rectangle having the size of the specified threshold range is vertical. It is determined whether there are rectangles of the same size in the direction, and if there are rectangles of the same size, row rectangles are generated by sequentially connecting them.

<Other embodiments>
Although the present embodiment shows a case where the present invention is applied to character determination processing as image processing before creating a PDF file, the present invention is not limited to this, and for example, an OCR (Optical Character Reader) The present invention can also be applied to character determination processing so that characters are not erroneously recognized as preprocessing of such character recognition processing.

  Furthermore, it is possible to provide a character determination program for causing a computer having the CPU 4 to execute character determination processing executed by the MFP 10 according to the present embodiment. Such a program is stored in a computer-readable recording medium such as a flexible disk attached to the computer, a CD-ROM (Compact Disk-Read Only Memory), a ROM (Read Only Memory), a RAM (Random Access Memory), and a memory card. And can be provided as a program product. Alternatively, the program can be provided by being recorded on a recording medium such as a hard disk built in the computer. A program can also be provided by downloading via a network.

  The character determination program according to the present invention causes information management processing to be executed by calling necessary modules at a predetermined timing in a predetermined arrangement among program modules provided as a part of a computer operation system (OS). It may be a thing. In that case, the program itself does not include the module, and the information management process is executed in cooperation with the OS. A program that does not include such a module can also be included in the character determination program according to the present invention.

The provided program product is installed in a program storage unit such as a hard disk and executed. The program product includes the program itself and a recording medium on which the program is recorded.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

FIG. 3 is a diagram showing a specific example of the hardware configuration of the MFP according to the present embodiment. It is a block diagram which shows the function structure for producing a PDF file. It is a functional block diagram which shows the function structure of a character area specific | specification part. It is an image figure which shows the pixel expansion | swelling to the main scanning direction. It is an image figure which shows pixel expansion to the subscanning direction of the image data after expansion to the right direction. It is an image figure which shows the pixel expansion to the subscanning direction of the image data after expansion to the left direction. It is an image figure which shows the union of the black pixel of the image data after expansion to the left-right direction. It is an image figure which shows the pixel expansion to the main scanning direction after rotation. It is an image figure which shows pixel expansion to the subscanning direction of the image data after expanding to the right after rotation. It is an image figure which shows the pixel expansion | swelling to the subscanning direction of the image data after expansion to the left direction after rotation. It is an image figure which shows the union of the black pixel of the image data after expanding in the left-right direction after rotation. It is an image figure which shows rotation of image data. It is an image figure which shows an expansion process. It is an image figure which shows the 1st connection process of 1st character line rectangle (black pixel). It is an image figure which shows the pixel data obtained by superimposing the pixel data which performed only the expansion process. It is an image figure which shows the judgment method of whether it is a character area in a specific part. It is a flowchart which shows the process sequence which produces a PDF file. It is a flowchart which shows the process sequence of a character area specific process. It is a figure which shows an example of the color image input. It is an image figure which shows a binary image. It is a flowchart which shows the process sequence of a blank area | region extraction process. It is a flowchart which shows the process sequence of an expansion process. It is an image figure which shows the image data of the state which acquired the union of the 1st row rectangle by superimposing the image after the expansion process before rotation, and the image after the expansion process after rotation. It is a flowchart which shows the process sequence of a character line direction determination process. It is a flowchart which shows the process sequence of a 1st connection process. It is a flowchart which shows the process sequence of a 1st character determination process. It is an image figure which shows the image after the 1st connection process to a character line direction. It is a flowchart which shows the process sequence of a 2nd connection process. It is a flowchart which shows the process sequence of a character color determination process. It is an image figure which shows the image of the final character area before compact PDF data creation.

Explanation of symbols

  1 scan processing unit, 2 input image processing unit, 3 storage unit, 4 CPU, 5 network I / F, 6 output image processing unit, 7 engine unit, 8 modem / NCU, 9 operation unit, 10 image processing device (MFP) , 11 generation unit, 12 first extraction unit, 13 expansion unit, 14 determination unit, 15 calculation unit, 16 first determination unit, 17 first connection unit, 18 first identification unit, 19 second extraction 20, second determination unit, 21 third determination unit, 22 second connection unit, 23 second identification unit, 101 image data acquisition unit, 103 preprocessing unit, 105 region identification unit, 107 lossless compression unit 109, a resolution reduction unit, 111 a lossy compression unit, and a 113 PDF conversion unit.

Claims (18)

An image processing device for specifying a character region from an image composed of a plurality of pixels,
Generating means for generating a binary image based on the image;
First extraction means for extracting a margin area and a non-margin area other than the margin area from the binary image;
Wherein by performing the expansion processing of the pixels in the first direction and a second direction perpendicular to it in a non-blank area, and expansion means for generating a first row rectangle, respectively,
Determining means for determining a character line direction of the binary image based on an aspect ratio of the first line rectangle;
First connecting means for generating a second line rectangle by connecting the first line rectangles in the character line direction determined by the determining means;
An image processing apparatus comprising: a first specifying unit that specifies a character area based on an area corresponding to the second row rectangle in the binary image. The determination means includes
Among the first row rectangles, the number of row rectangles satisfying a first predetermined condition indicating that the aspect ratio is a horizontally long row rectangle, and the horizontally long row rectangles having a long aspect ratio. Calculating means for calculating the number of row rectangles satisfying a second predetermined condition indicating that;
The image processing apparatus according to claim 1, further comprising: a first determination unit that determines a character line direction of the binary image based on the two numbers. The first specifying means includes:
Second extraction means for extracting at least one small rectangle included in the second row rectangle;
For each second row rectangle, when the ratio of the number of small rectangles to the aspect ratio of the second row rectangle is within a predetermined range, an area corresponding to the second row rectangle is determined as a character area. The image processing apparatus according to claim 1, further comprising: a second determination unit that performs the determination. The first specifying means includes:
Second extraction means for extracting at least one small rectangle included in the second row rectangle;
For each of the second row rectangles, when the variation in the size of the small rectangles included in the second row rectangle is equal to or less than a predetermined value, a region corresponding to the second row rectangle is determined as a character region. The image processing apparatus according to claim 1, further comprising: 3 determination means. A third row rectangle is obtained by connecting the first row rectangles other than the region identified as the character region by the first identifying unit in a direction perpendicular to the character row direction determined by the determining unit. Second connecting means to generate;
5. The image processing apparatus according to claim 1, further comprising: a second specifying unit that specifies a character area based on an area corresponding to a third row rectangle in the binary image. 6. The generation means generates a plurality of types of binary images corresponding to each of a plurality of colors obtained by performing a color reduction process on the image,
The first extraction means, the expansion means, the determination means, and the first connection means perform processing on each binary image,
6. The device according to claim 1, wherein the first specifying unit specifies a character region based on a union of regions corresponding to the second row rectangles in the respective binary images. Image processing device. A character area specifying method using an image processing device for specifying a character area from an image composed of a plurality of pixels,
The image processing apparatus includes:
With a control unit,
The image processing method includes:
The control unit generating a binary image based on the image;
The control unit extracting a margin area and a non-margin area other than the margin area from the binary image;
A step wherein said control unit is for generating each said by performing an expansion process in the first direction and the pixel in a second direction perpendicular thereto in a non-blank area, the first row rectangle,
The control unit determining a character line direction of the binary image based on an aspect ratio of the first line rectangle;
The controller generates a second line rectangle by connecting the first line rectangles in the character line direction; and
And a step of specifying the character area based on an area corresponding to the second row rectangle in the binary image. The step of determining includes
Among the first row rectangles, the number of row rectangles satisfying a first predetermined condition indicating that the aspect ratio is a horizontally long row rectangle, and indicating that the aspect ratio is a vertically long row rectangle. Calculating the number of row rectangles satisfying the second predetermined condition;
The character region specifying method according to claim 7, further comprising: determining a character line direction of the binary image based on the two numbers. The step of specifying the character area includes:
Extracting at least one small rectangle included in the second row rectangle;
For each second row rectangle, when the ratio of the number of small rectangles to the aspect ratio of the second row rectangle is within a predetermined range, an area corresponding to the second row rectangle is determined as a character area. The character area specifying method according to claim 7 or 8, comprising the step of: The step of specifying the character area includes:
Extracting at least one small rectangle included in the second row rectangle;
Determining a region corresponding to the second row rectangle as a character region when variation in size of small rectangles included in the second row rectangle is equal to or less than a predetermined value for each of the second row rectangles; The character area specifying method according to claim 7, comprising: A third row is formed by connecting the first row rectangles other than the region identified as the character region by the step of identifying the character region in a direction perpendicular to the character row direction determined in the determination step. Generating a rectangle;
The character area specifying method according to claim 7, further comprising: specifying a character area based on an area corresponding to a third row rectangle in the binary image. In the generating step of generating the binary image, a plurality of types of binary images corresponding to each of a plurality of colors obtained by performing a color reduction process on the image are generated,
The steps of extracting the margin area and other non-margin areas, generating the first row rectangle, determining, and generating the second row rectangle, Processing the binary image;
12. The character region is specified based on a union of regions corresponding to the second row rectangles in the respective binary images in the step of specifying the character region. 12. Character area identification method. A character area specifying program for causing a computer to specify a character area from an image composed of a plurality of pixels,
The program is stored in the computer.
Generating a binary image based on the image;
Extracting a margin area and other non-margin areas from the binary image;
Wherein by performing the expansion processing of the pixels in the first direction and a second direction perpendicular to it in a non-blank area, and generating a first row rectangle, respectively,
Determining a text line direction of the binary image based on an aspect ratio of the first line rectangle;
Generating a second line rectangle by connecting the first line rectangles in the character line direction;
And a step of specifying a character area based on an area corresponding to the second row rectangle in the binary image. The step of determining includes
Among the first row rectangles, the number of row rectangles satisfying a first predetermined condition indicating that the aspect ratio is a horizontally long row rectangle, and indicating that the aspect ratio is a vertically long row rectangle. Calculating the number of row rectangles satisfying the second predetermined condition;
The character area specifying program according to claim 13, further comprising: determining a character line direction of the binary image based on the two numbers. The step of specifying the character area includes:
Extracting at least one small rectangle included in the second row rectangle;
For each second row rectangle, when the ratio of the number of small rectangles to the aspect ratio of the second row rectangle is within a predetermined range, an area corresponding to the second row rectangle is determined as a character area. 15. The character area specifying program according to claim 13 or 14, comprising the step of: The step of specifying the character area includes:
Extracting at least one small rectangle included in the second row rectangle;
Determining a region corresponding to the second row rectangle as a character region when variation in size of small rectangles included in the second row rectangle is equal to or less than a predetermined value for each of the second row rectangles; The character area specifying program according to claim 13, comprising: A third row is formed by connecting the first row rectangles other than the region identified as the character region by the step of identifying the character region in a direction perpendicular to the character row direction determined in the determination step. Generating a rectangle;
The character area specifying program according to any one of claims 13 to 16, further causing the computer to execute a step of specifying a character area based on an area corresponding to a third row rectangle in the binary image. In the generating step of generating the binary image, a plurality of types of binary images corresponding to each of a plurality of colors obtained by performing a color reduction process on the image are generated,
The steps of extracting the margin area and other non-margin areas, generating the first row rectangle, determining, and generating the second row rectangle, Processing the binary image;
18. The character area is specified based on a union of areas corresponding to the second row rectangles in each of the binary images in the step of specifying the character area. 18. Character area identification program.