JP2010011450A - Image-forming device and image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image-forming device capable of reducing the risk of deteriorating quality of an output image by reducing an increase in image area identification processing time depending on an input document image. <P>SOLUTION: The image-forming device has: an image input means 101 for acquiring image information from an original to generate an input image; a first identification means 102 for performing first image area identification processing to the input image to output a first identification signal indicating the attribute of an image; a second identification means 104 for inputting the first identification signal and performing second image area identification processing following the first image area identification processing to output a second identification signal; a determination means 103 for inputting the first identification signal to output a determination signal indicating whether the second identification means should be executed; a selection means for selecting the first identification signal as a third one without performing the second image area identification processing when the determination signal does not indicate the execution and for selecting the second identification signal as the third one when the determination signal indicates the execution; and an image processing means 105 for inputting the input image and the third identification signal to execute image processing without performing the second image area identification processing. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an image forming apparatus and an image processing method, and more particularly to a technique for performing image area identification processing on a document image to form an image.

  A technique is known in which image area identification processing is performed on a document image, and image formation is performed by switching image processing to be applied based on the identification result. Recently, in order to reduce the amount of data while maintaining the image quality, text and photos are separated, text emphasizes resolution, and photos emphasize gradation and image processing suitable for each.・ There is a technology to compress files into electronic files.

  For example, in the technique described in Patent Document 1, a layout analysis process is applied to image data that is a processing target, an image element of a character part is specified based on the result, an image element specified as a character part, By compressing the remaining portions with different compression methods, a highly compressed image file (highly compressed PDF) is obtained while maintaining character readability.

For the above-described image area identification processing, techniques disclosed in Patent Document 2 and Patent Document 3 are known.
In the technique described in Patent Document 2, after binarizing the processing target, it is determined whether or not the black pixel is a character based only on the size of the circumscribed rectangle of the connected component of the black pixel group.
In the technique described in Patent Document 3, a character string and a character string region are extracted based on the size and arrangement of a circumscribed rectangle set of a specific pixel group.

  With the technique described in Patent Document 2, misidentification is unlikely to occur and processing can be performed at a high speed, but applicable objects are narrow such that characters in a photograph cannot be extracted. Therefore, when the above-described highly compressed image file is formed using this identification result, characters other than black characters are blurred and readability deteriorates, and it is difficult to obtain a high compression effect. .

  On the other hand, in the technique described in Patent Document 3, it is possible to extract various characters compared to the former. However, when a document image including many halftone dot backgrounds or a document image having a complicated configuration is input, the number of circumscribed rectangles becomes enormous and the layout analysis process may take a very long time. In addition, there is a high risk of erroneous identification in the layout analysis process. Regarding the point that requires a long processing time, there is a saving method in which the identification processing is performed only within the predetermined processing time and the identification processing is not performed when the processing time is exceeded, but as a result, the image desired by the user is formed. Can not do it.

  The present invention has been made in view of such circumstances, and reduces an increase in image area identification processing time that depends on an input document image, reduces the risk of image quality degradation of an output image, and reduces user expectations. It is an object of the present invention to provide an image forming apparatus and an image processing method capable of obtaining an output image close to.

  In order to solve the above problems, the present invention provides an image input means for acquiring image information from a document including a paper document and generating an input image, and performing a first image area identification process on the input image to obtain an image First identification means for outputting a first identification signal indicating an attribute, and the first identification signal are input, and a second image area identification process following the first image area identification process is performed to perform image identification. Second identification means for outputting a second identification signal indicating an attribute; and determination means for inputting the first identification signal and outputting a determination signal indicating whether or not to execute the second identification means; When the determination signal indicates that it should not be executed, the first identification signal is selected as a third identification signal, and when the determination signal indicates that it should be executed, the second identification signal is selected. Selecting means for selecting an identification signal as the third identification signal; Enter a third identification signal which is an image forming apparatus and an image processing means for executing image processing.

  According to another aspect of the present invention, there is provided an image input unit that obtains image information from a document including a paper document and generates an input image, and a first image area identification process performed on the input image to indicate an attribute of the image. A first identification means for outputting an identification signal; and the first identification signal is inputted, a preprocessing of a second image area identification process following the first image area identification process is performed, and a preprocessed signal is output Pre-processing means for inputting, pre-processing signals, second identifying means for performing the second image area identification processing and outputting second identification signals indicating image attributes, and pre-processing signals are inputted. Determining means for outputting a determination signal indicating whether or not the second identification means should be executed; and if the determination signal indicates that the second identification means should not be executed, If it is selected as an identification signal and the decision signal indicates that it should be executed, the second identification signal Selection means for selecting as said third identification signals, and inputs the said third identification signal and the input image is an image forming apparatus and an image processing means for executing image processing.

  According to another aspect of the present invention, there is provided an image input unit that obtains image information from a document including a paper document and generates an input image, and a first image area identification process performed on the input image to indicate an attribute of the image. A first identification unit that outputs an identification signal; a second identification unit that inputs the first identification signal and performs a second image area identification process subsequent to the first image area identification process; A second identification unit that outputs an identification signal; a determination unit that inputs the first identification signal and outputs a determination signal indicating whether or not the second identification unit should be executed; and the determination signal is executed When it is indicated that it should not be performed, it indicates that the input signal and the first identification signal are input to execute image processing, and that the determination signal is to be executed. In the case, the input image and the second identification signal are input to execute image processing. An image forming apparatus and a second image processing means.

  The present invention also acquires image information from a document including a paper document to generate an input image, and performs a first image area identification process on the input image to generate a first identification signal indicating an attribute of the image. Generating a second identification signal indicating an image attribute by performing a second image area identification process subsequent to the first image area identification process on the first identification signal, and generating the first identification signal A determination signal indicating whether or not the second identification means should be executed is generated from the signal, and when the determination signal indicates that the second identification means should not be executed, the second image area identification processing is not performed. When the first identification signal is selected as a third identification signal and the determination signal indicates that it should be executed, the second identification signal is selected as the third identification signal, and the input image An image for which image processing is performed without performing the second image area identification processing on the third identification signal. It is a processing method.

  According to the image forming apparatus and the image processing method of the present invention, the increase in the image area identification processing time depending on the input document image is reduced, and the risk of the image quality deterioration of the output image is reduced, which is close to the user's expectation. An output image can be obtained.

1 is a diagram illustrating a configuration of an image forming apparatus according to a first embodiment. 3 is a flowchart showing a schematic processing procedure of the image forming apparatus according to the first embodiment. The figure which shows the data format of the output file of 1st Embodiment. 1 is a diagram illustrating a conventional image forming apparatus. The figure explaining a labeling process. The figure which illustrates the circumscribed rectangle and character component of the connection component of an edge image. The block diagram which shows the example of 1 structure of an image processing means. FIG. 4 is a diagram illustrating a configuration of an image forming apparatus according to a second embodiment. 9 is a flowchart illustrating a schematic processing procedure of the image forming apparatus according to the second embodiment. FIG. 9 is a diagram illustrating a configuration of an image forming apparatus according to a third embodiment. 10 is a flowchart illustrating a schematic processing procedure of the image forming apparatus according to the third embodiment. The figure explaining the data format of the PDF file produced | generated by the 1st image processing means of 3rd Embodiment. The block diagram which shows the example of 1 structure of a 1st image processing means. The figure which compares and compares the data size of the image of a MRC format and another format. The figure which shows the operation example of a character color detection means. The figure explaining the calculation method of a projection image. The figure which classifies and shows the function of the 1st and 2nd identification means.

[First embodiment]
FIG. 1 is a diagram illustrating a configuration of the image forming apparatus according to the first embodiment, and FIG. 2 is a flowchart illustrating a schematic processing procedure of the image forming apparatus according to the first embodiment. The configuration and operation of the image forming apparatus will be described with reference to FIGS.

  The image input unit 101 is, for example, an MFP scanner that converts an input paper document into an image signal. In Act 201, the image input means 101 outputs the converted input image signal 111. In Act 202, the first identification means 102 inputs the input image signal 111, performs a first identification process with reference to pixels in a relatively narrow range of the input image by edge extraction / black pixel extraction, etc. A first identification signal 112 indicating a pixel attribute is output. In Act 203, the determination unit 103 receives the first identification signal 112, determines the complexity of the input document image, and outputs a determination signal 113 indicating whether the input document image is complicated.

  In Act 204, the second identification means 104 receives the first identification signal 112, executes the second identification process, extracts the character string / character region, and outputs the second identification signal 114. However, if the determination signal 113 indicates that “the input document is complicated”, the second identification means does not perform processing, and “the input document is not complicated (standard)”. Is indicated, the second identification means performs processing.

  In Act 205, the image processing means 105 inputs the input image signal 111 and the first identification signal 112 or the second identification signal 114 and performs image processing. Here, when the determination signal 113 indicates that “the input document is complicated”, the first identification signal 112 is input as the identification signal. When the determination signal 113 indicates that “the input document is not complicated (standard)”, the second identification signal 114 is input as the identification signal.

  In addition, the well-known technique can be used for the 1st identification means 102 and the 2nd identification means 104 in 1st Embodiment. For example, the same processing as that of the character region extraction unit described in Japanese Patent Laid-Open No. 2000-20726 can be used.

Next, the first embodiment will be specifically described.
In the first embodiment, an MFP (multifunction multi-function peripheral) is taken as an example of an image forming apparatus, and high compression PDF generation that is mounted as a standard as a scanning function of the MFP will be described.

FIG. 3 is a diagram illustrating a data format of the output file according to the first embodiment.
Even in an electronic file generally called a highly compressed PDF, there are various data formats in the file. In the high compression PDF of the first embodiment, high compression is realized by cutting out an object such as a character / natural image from the input image 301 shown in FIG. 3 and compressing and combining them in a method suitable for the attribute.

  That is, in this data format, a 300 dpi binary character image 303 divided for each color is superimposed on the 150 dpi background image 302 (not shown in the figure). Here, the character image 303 has binary values of “0” and “1”. In the character image 303, “0” value pixels corresponding to white pixels are drawn through the background image 302 in the layer below the overlap, and “1” value pixels corresponding to black pixels are specified separately. Draw in the color that is present. The character image is clearly represented because it is expressed at 300 dpi, and the resolution is reduced to 150 dpi for portions other than the characters, so that the file size can be reduced.

  Here, for comparison, the configuration and operation of a conventional image forming apparatus will be described. FIG. 4 is a diagram illustrating a conventional image forming apparatus.

  The image input unit 401 outputs an input image signal 411. The identification unit 402 outputs an identification signal 412 based on the input image signal 411. The image processing unit 406 receives the input image signal 411 and the identification signal 412 and generates and outputs a highly compressed PDF file.

  Here, the identification unit 402 includes an edge extraction unit 403, an outer rectangle generation unit (labeling unit) 404, and a character string / character area extraction processing unit 405.

  The edge extraction unit 403 receives the input image signal 411 and outputs an edge image obtained by extracting edge pixels in the input image. Various methods have been proposed as edge extraction means. For example, after obtaining the luminance value Y by the conversion equation shown in equation (1), extraction is performed using a simple Sobel filter shown in the lower matrix.

Y = 0.257R + 0.504G + 0.098B + 16 Formula (1)
(Where R, G and B are signal values of each pixel)

  Outer rectangle generation means (labeling) 404 receives the edge image output from the edge extraction means 403, and outputs an outer rectangle image in which a circumscribed rectangle of a connected component of edge pixels is generated. In order to generate a circumscribed rectangle, an outer rectangle generation unit (labeling) 404 executes a labeling process.

FIG. 5 is a diagram illustrating an example of the labeling process. Various labeling methods have been proposed, and it goes without saying that any of them may be selected.
The input image 501 is a binary image of “0” and “1”. The input image 501 is labeled by scanning all pixels in the order of left → right, top → bottom, starting from the top left pixel. Now, assuming that a pixel to be scanned is a pixel of interest, a label (number) attached to the pixel of interest is determined according to the situation of the corresponding reference pixel. In the processing method defined in the example 502 indicating the target pixel and the reference pixel, when the target pixel (O) is a black pixel, the reference pixel (*) at the left, upper left, upper, and upper right positions of the target pixel is displayed. The label of the pixel of interest is determined according to the value.

For example, label with the following rules:
* If there is no black pixel in the reference pixel, numbers are numbered in ascending order starting from 0.
* If there is one type of label in the reference pixel, attach the same label as in the reference pixel.
* If there are two or more types of labels in the reference pixel, the label with the lowest number in the reference pixel is attached. Furthermore, it is recorded that the labels with the different numbers are the same label.

  If the pixel in the upper left (1 row × 1 column) is the target pixel, the reference pixel is outside the region and there is no black pixel. Therefore, a new label “0” is attached. Next, if the pixel on the right side (1 row × 2 columns) is the target pixel, the reference pixel of 1 row × 1 column is a black pixel. Therefore, the same label “0” as that in the reference pixel (1 row × 1 column) is attached.

  Further, the pixel of (1 row × 3 columns) and the pixel of (1 row × 4 columns) proceeding to the right side are white pixels. Therefore, no label is given as a blank for this pixel of interest. If the pixel in the upper right (1 row × 5 columns) is the target pixel, there is no black pixel as the reference pixel. Therefore, a new label “1” in ascending order is attached.

  If the pixel of (4 rows × 4 columns) is the target pixel, the reference pixel on the left side of the target pixel is already labeled “0”, while the upper reference pixel is labeled “1”. Yes. Accordingly, the pixel of interest is labeled “0”, which is a smaller number among “0” and “1”. Further, it is recorded that “0” and “1” are the same label.

  If the pixel in the lower left (5 rows × 1 column) is the target pixel, no black pixel exists in the reference pixel. Therefore, add a new label. At this time, since labels “0” and “1” are already used, a new label “2” is attached.

  The labeling result 503 shows the state of the label after the first scan is completed in this way. From the labeling result 503, it can be seen that the connected components that should originally have the same label are divided into two groups of pixels having the label “0” and pixels having the label “1”.

  Next, label unification processing is performed. As described above, when the fourth pixel from the left and the fourth pixel from the top are the target pixels, it is recorded separately that “0” and “1” are the same label. In such a case, unify labels with young numbers. That is, a pixel labeled “1” in the labeling result 503 is relabeled “0”. This processing result is shown as a labeling result 504. As shown in the labeling result 504, all the pixels of the group that should originally be one connected component have the same label.

  Finally, empty label correction processing is executed. If an empty label is generated as a result of the unification process, the label number is repacked. In the example of the labeling result 504, since the label “1” becomes an empty number, the label numbers after “2” are packed. A labeling result 505 indicates a labeling result finally obtained in this manner.

  Since the set of the minimum value and the maximum value of the (x coordinate, y coordinate) of the pixel with the same label are respectively the upper left vertex and the lower right vertex (x coordinate, y coordinate) of the circumscribed rectangle of the connected component. The circumscribed rectangle of the connected component of the edge pixel can be generated from the result of the labeling process.

  The character string / character area extraction processing unit 405 receives the outer rectangular image output from the outer rectangular generation unit 404 and outputs an identification signal 412. FIG. 6 is a diagram illustrating the circumscribed rectangle and the character component of the connected component of the edge image.

Specifically, the character string / character region extraction processing unit 405 receives a circumscribed rectangular image 601 as a connected component of the edge image (a dashed-dotted line in the drawing is a circumscribed rectangle), and is a rectangle that is in contact, intersecting, or included. The character components 602 are generated by combining them. Then, the character string / character area extraction processing unit 405 does not have an extremely large or small size for the character component generated in this way.
• Is the aspect ratio of the rectangle close to 1 (is the shape of the rectangle close to a square)?
・ Is the character component and size in the horizontal direction aligned? Are the y coordinates aligned?
・ Is the size of the character component nearby in the vertical direction aligned? Are the x coordinates aligned?
A character string is generated based on whether or not the character string has a characteristic such as a character string, and it is further determined whether or not it is a character region from the character string characteristic and the character string arrangement method.

  Here, in the process of extracting the character string / character region, the deletion of the rectangle and the combination of the rectangles are repeatedly executed. Therefore, when there are many character components to be processed, the processing time becomes enormous. In addition, the fact that there are many character components also indicates that there is a high possibility that the identification accuracy is poor because there are many halftone dots and noise that are easily misidentified as characters.

  The image processing unit 406 will be described. FIG. 7 is a block diagram showing an example of the configuration of the image processing means. The image processing means 406 has a configuration for obtaining an image file having the data format shown in FIG.

  The character pixel detection unit 701 performs image area identification processing based on the input image signal and outputs an identification signal. That is, the character pixel detection unit 701 corresponds to the identification unit 402 in FIG.

  First, the process of creating the background multi-value image 302 will be described. The character surrounding color detection unit 702 detects the color around the character. As a detection method, for example, pixels within a distance of 3 pixels from a character are extracted, and an average value of the pixel values is obtained as a character surrounding color. Next, the character filling means 703 replaces the character pixel with the detected character surrounding color. This fills the character with the surrounding background color. Finally, the image reduction means 704 executes reduction processing, and then the multi-value image compression means 705 executes multi-value image compression such as JPEG to create a background image.

  It should be noted that the character filling means 703 paints characters because multi-valued image compression represented by JPEG generally has a tendency not to compress images with many high-frequency components of the spatial frequency, and therefore erases characters containing high-frequency components. This is to increase the compression ratio.

  Next, a process for creating the character image 303 will be described. First, the character color detection means 706 detects the character color of each character pixel. Specifically, for example, character pixels of similar colors are grouped, and the RGB average value of the character pixels of the group is set as the character color. On the other hand, in parallel with this processing, the binarizing means 707 generates a binary image so that the character pixel is “1” and the non-character pixel is “0”.

A binary image dividing unit 708 by character color divides the generated binary image into binary images for each color grouped by the character color detecting unit 706. A binary image compression unit 709 performs binary image compression such as MMR on the divided binary image to create a character image.
Note that the reason why the binary image is divided by grouping the character pixels for each color is that it is a binary image, and therefore only one color of the background and the foreground (character) can be expressed in one image.

  The configuration and operation of the conventional image forming apparatus have been described above. As described above, when the input document image is complicated in the conventional image forming apparatus, there is a risk that the processing time in the image area identification unit becomes enormous, or the image quality defect occurs due to insufficient identification accuracy. high.

  The image forming apparatus according to the first embodiment shown in FIG. 1 prevents such problems from occurring. The configuration and operation of the image forming apparatus according to the first embodiment will be described in comparison with a conventional image forming apparatus with reference to FIGS.

  In the image forming apparatus according to the first embodiment, the identification unit 402 (FIG. 4) of the conventional image forming apparatus is replaced with the first identification unit 102 (FIG. 1) and the second identification unit 104 (FIG. 1). To divide. Here, the first identification unit 102 (FIG. 1) corresponds to the edge extraction unit 403 (FIG. 4) of the conventional image forming apparatus, and the second identification unit 104 generates the outer rectangle of the conventional image forming apparatus. It corresponds to the means 404 (FIG. 4) and the character string / character area extraction processing means 405 (FIG. 4).

  In the image forming apparatus according to the first embodiment, the determination unit 103 (FIG. 1) determines the document complexity. That is, the determination means 103 (FIG. 1) receives the first identification signal 112 (FIG. 1) and outputs a determination signal 113 (FIG. 1) as to whether or not the document is complicated.

  Here, the document image complexity determination unit 103 (FIG. 1) will be described. As described above, the second identification unit 104, that is, the outer rectangle generation unit 404 (FIG. 4) and the character string / character area extraction processing unit 405 (FIG. 4) of the conventional image forming apparatus, is the number of circumscribed rectangles of the edge image. The processing time becomes enormous or the identification accuracy is likely to deteriorate. Here, the number of circumscribed rectangles of edge pixels is considered to depend to some extent on the number of edge pixels.

  Therefore, the determination unit 103 (FIG. 1) uses the edge image signal output from the first identification unit 102, that is, the edge extraction unit 403 (FIG. 4) of the conventional image forming apparatus, as the first identification signal 112 (FIG. 1). ) And the edge image is scanned to count the number of edge pixels Ne. If Ne> Th1, a determination signal 113 indicating that the input document is complicated is output. On the other hand, if Ne <= Th1, a determination signal 113 indicating that the input document is not complicated (standard) is output. Here, Th1 is a predetermined threshold value.

  When the determination signal 113 (FIG. 1) indicates that the document is complicated, the second identification unit 104, that is, the outer rectangle generation unit 404 (FIG. 4) of the conventional image forming apparatus and the character The column / character area extraction processing unit 405 (FIG. 4) does not operate. Accordingly, the image processing means 105 (FIG. 1) receives the input image signal 111 (FIG. 1) and the first identification signal 112 and outputs a highly compressed PDF.

  On the other hand, when the determination signal 113 (FIG. 1) indicates “the document is not complicated (standard)”, the second identification unit 104, that is, the outer rectangle of the conventional image forming apparatus is used. The generation means 404 (FIG. 4) and the character string / character area extraction processing means 405 (FIG. 4) operate. Therefore, the image processing means 105 (FIG. 1) receives the input image signal 111 (FIG. 1) and the second identification signal 114 as an input and outputs a highly compressed PDF.

As described above, according to the first embodiment, it is possible to create a highly compressed PDF file with less risk of image quality problems and enormous processing time even when the input document image is complicated.
Further, the document image complexity determination unit 103 according to the first embodiment uses only data necessary for the second identification unit 104 and does not require new data input. Therefore, the determination can be made without greatly affecting the processing time. Furthermore, when the input document image is complicated, it is possible to create a highly compressed PDF without generating unnecessary signals as compared with the conventional method.

[Second Embodiment]
FIG. 8 is a diagram illustrating a configuration of the image forming apparatus according to the second embodiment, and FIG. 9 is a flowchart illustrating a schematic processing procedure of the image forming apparatus according to the second embodiment. The configuration and operation of the image forming apparatus will be described with reference to FIGS.
The image forming apparatus according to the second embodiment is different from the image forming apparatus according to the first embodiment in that the image forming apparatus according to the second embodiment includes a pre-processing unit 803 for second identification.

  The image input means 801 is an MFP scanner or the like that converts an input paper document into an image signal. In Act 901, the image input means 801 outputs the converted input image signal 811. In Act 902, the first identification means 802 receives the input image signal 811 and executes a first identification process with reference to pixels in a relatively narrow range of the input image by edge extraction / black pixel extraction, etc. A first identification signal 812 indicating a pixel attribute is output. In Act 903, the second identification preprocessing means 803 receives the first identification signal 812, performs preprocessing of the second identification means, and outputs a preprocessing signal 813.

  In Act 904, the determination unit 804 receives the preprocess signal 813, determines the complexity of the input document image, and outputs a determination signal 814 indicating whether the input document image is complicated. In Act 905, the second identification unit 805 receives the preprocess signal 813, extracts a character string / character region, and outputs a second identification signal 815. However, when the determination signal 814 indicates that “the input document is complicated”, the second identification unit 805 does not perform processing, and “the input document is not complicated (standard)”. Is indicated, the second identification means 805 performs processing.

  In Act 906, the image processing means 806 performs image processing with the input image signal 811 and the first identification signal 812 or the second identification signal 815 as inputs. Here, when the determination signal 814 indicates that “the input document is complicated”, the first identification signal 812 is input as the identification signal. When the determination signal 814 indicates “input document is not complicated (standard)”, the second identification signal 815 is input as the identification signal.

  The major difference between the second embodiment and the first embodiment is that in the first embodiment, the identification means 402 (FIG. 4) of the conventional image forming apparatus is replaced with the first identification means 102 (FIG. 1). ) And the second identification means 104 (FIG. 1), whereas in the second embodiment, the first identification means 802 (FIG. 8) and the second identification preprocessing means are divided. It is divided into three parts 803 (FIG. 8) and second identification means 805 (FIG. 8).

The specific correspondence between the units of the second embodiment and the processing blocks of the conventional image forming apparatus is as follows.
The first identification unit 802 of the second embodiment corresponds to the edge extraction unit 403 (FIG. 4) of the prior art. The second identification pre-processing means 803 of the second embodiment corresponds to the prior art outer rectangle generation means 404 (FIG. 4). The second identification means 805 of the second embodiment corresponds to the conventional character string / character area extraction processing means 405 (FIG. 4).

  Here, the document image complexity determination means 804 (FIG. 8) will be described. As described above, the conventional character string / character area extraction processing unit 405 (FIG. 4) corresponding to the second identification unit 805 recognizes that the processing time becomes enormous depending on the number of circumscribed rectangles of the edge image. The accuracy may deteriorate. In the first embodiment, the number of edge pixels Ne is counted, and whether Ne> Th is used as a criterion for determining whether the input document is complicated.

In the second embodiment, since the pre-processing unit 803 for second identification outputs an outer rectangular image as the pre-processing signal 813, the determining unit 804 uses the outer rectangular number Nr that more represents the complexity of the document. If Nr> Th2 is counted, it is determined that the input document is complex, and a determination signal 814 indicating that the input document is complex is output. On the other hand, if Nr <= Th2, a determination signal 814 indicating that the input document is not complicated (standard) is output. Here, Th2 is a predetermined threshold value.
This makes it possible to determine the complexity of the document more accurately than in the first embodiment.

  As described above, in the second embodiment, it is possible to determine the complexity of the input document image more accurately than in the first embodiment. Even when the input document image is complicated, image quality defects and It is possible to create a highly-compressed PDF file with less risk of processing time becoming enormous.

  The document image complexity determination unit 804 of the second embodiment uses only data necessary for the second identification unit 805 and does not require new data input. Therefore, the determination can be made without greatly affecting the processing time.

[Third embodiment]
FIG. 10 is a diagram illustrating a configuration of the image forming apparatus according to the third embodiment, and FIG. 11 is a flowchart illustrating a schematic processing procedure of the image forming apparatus according to the third embodiment. The configuration and operation of the image forming apparatus will be described with reference to FIGS.
Note that the difference between the third embodiment and the first embodiment is that the image processing means is switched depending on the complexity of the input document image. In the third embodiment, the first image processing means 1005 is switched. And second image processing means 1006.

  In the third embodiment, if it is determined in Act 1103 that the input image is complicated by the complexity determining unit 1003, the first image processing unit is based on the first identification signal 1012 in Act 1106. In step 1005, the first image processing is performed. On the other hand, if it is determined that the input image is not complicated (standard), in Act 1104, the second identification unit 1004 outputs the second identification signal 1013 based on the first identification signal 1012. In Act 1105, second image processing is performed by the second image processing unit 1006 based on the second identification signal 1013.

  FIG. 12 is a diagram for explaining the data format of the PDF file generated by the first image processing unit 1005 of the third embodiment.

  The data format (MRC) shown in FIG. 12 includes a mask image 1202, a background image 1203, and a character color image 1204. The mask image 1202 is a 300 dpi image having binary values of “0” and “1”. The background image 1203 is a 150 dpi image obtained by deleting characters from the input image. The character color image 1204 is a 75 dpi image for specifying the character color.

In the third embodiment, the mask image 1202 selects and displays the character color image 1204 at the position of the black pixel and the background image 1203 at the position of the white pixel. Thereby, since the shape of the character depends on the 300 dpi mask image, the file size is reduced by reducing the resolution other than the character while keeping the character clear.
The data format of the PDF file created by the second image processing unit 1006 of the third embodiment is shown in FIG. 3 generated by the image processing unit 105 (FIG. 1) of the first embodiment. The data format is the same as that of the PDF file.

  FIG. 13 is a block diagram illustrating a configuration example of the first image processing unit 1005.

  The character pixel detection unit 1301 performs image area identification processing based on the input image signal and outputs an identification signal. That is, since the character pixel detection unit 1301 corresponds to the first identification unit 1002 of FIG. 10, detailed description thereof is omitted.

  First, the creation operation of the mask image 1202 will be described. For example, the binarizing unit 1302 generates a binary image in which character pixels are “1” and pixels other than characters are “0”. The binary image compressing unit 1303 creates a mask image 1202 by performing binary compression such as MMR on the generated binary image.

  Next, the creation operation of the background image 1203 will be described. Character surrounding color detection means 1304 detects the color around the character. As a detection method, for example, pixels within a distance of 3 pixels from a character are extracted, and an average value of the pixel values is obtained as a character surrounding color. Next, the character filling means 1305 replaces the character pixel with the detected character surrounding color. This fills the character with the surrounding background color. Finally, the first image reduction unit 1306 executes a reduction process, and then the first multi-value image compression unit 705 executes multi-value image compression such as JPEG to create a background image.

  Note that the character filling means 1305 paints characters because multi-valued image compression represented by JPEG generally has a tendency to be poor at compression of images having many high-frequency components of the spatial frequency. This is to increase the compression rate by erasing.

  Next, the creation operation of the character color image 1204 will be described. First, the non-character erasing unit 1308 erases characters other than characters from the input image. Specifically, pixels that are not detected as characters by the character detection unit 1301 are replaced with white pixels. Next, the character pixel expansion means 1309 expands the character pixel. Specifically, the process of replacing white pixels in the vicinity of the color pixels with that color is repeated for all color pixels. This character pixel expansion process is also a process for leaving no character shape in order to eliminate high-frequency components, as with the background image 1203. Therefore, if the number of repetitions is increased, the processing time increases, but the compressed file size can be reduced.

  Note that if the expansion process is repeated a certain number of times, the file size does not decrease, but the number of times depends on the input resolution. For example, the lower the resolution, the less the file size can be reduced with a smaller number of repetitions. The second image reduction unit 1310 performs a reduction process on the image after the character pixel expansion, and then the second multi-value image compression unit 1311 executes multi-value image compression to create a character color image 1204. To do.

Next, the reason why the data format is the MRC format will be described.
As shown in FIG. 12, in the MRC format output by the first image processing means 1005, a 150 dpi multi-value image for background, a binary image of 300 dpi for character shape, and a multi-value of 75 dpi for character color. It consists of images. On the other hand, as shown in FIG. 3, the format output by the second image processing means 1006 is composed of a 150 dpi multi-value image for background and a 300 dpi binary image for characters.

  Comparing the data sizes of the images constituting each of them, as shown in FIG. 14, the former (MRC format) and the latter background images have substantially the same data size. The former (MRC format) mask image and the latter character image have substantially the same data size. Therefore, the file size tends to be larger in the former (MRC format) by the data size of the character color image. However, the latter format requires processing with higher accuracy than the MRC format. Specifically, the character color detection means 706 (FIG. 7) required only in the latter format is required to have high accuracy.

FIG. 15 is a diagram showing an operation example of the character color detection means.
A character image 1501 in FIG. 15 is an example of an image input to the character color detection unit 706. The first line (alphabet), second line (number), and third line (Hiragana) of the character image 1501 are all different character colors.

  When grouping for each character color, if there is a one-to-one correspondence between the character color and the group as shown in the character image 1502 and the color of the group can be extracted correctly, as shown in the character image 1503, in the output electronic image file Can be expressed in substantially the same manner as the input image. However, for example, when a plurality of colors of characters in the first to third lines are mistakenly made into one character color group as shown in a character image 1504, as shown in a character image 1505, in the output electronic image file The character image is expressed by only one color (in this example, only one color of the average color of three lines).

In addition, when grouping is performed for each character as shown in the character image 1506, or when the extracted character color is mistaken for only one character in the character string as shown in the character image 1507, a character string of the same color There is also a risk that characters of different colors are generated in the image and an image with a very uncomfortable feeling is output.
The image data size itself is almost the same as that of the character images 1503, 1505, and 1507. However, if the image data size is too finely divided like the character image 1506, the overhead for constructing the output electronic image file becomes large. Can be large.

  That is, when the second image processing unit 1006 is executed when the input image is complicated, there is a high risk of wrong character color. Therefore, in the third embodiment, the first image processing unit 1005 uses the MRC format PDF. I try to create a file.

  As described above, according to the third embodiment, it is possible to create a highly compressed PDF file with less risk of image quality problems and enormous processing time even when the input document image is complicated. It is. Furthermore, the document image complexity determination process in the third embodiment uses only data necessary for the image area identification process, and does not require new data input. Therefore, determination can be made without adversely affecting the processing time when a standard image is input.

In the first to third embodiments, a high-compression PDF is taken as a specific example, and an image processing technique capable of preventing problems such as a large amount of processing time due to a complicated document and image quality defects. explained. In the following embodiment, an example of application of the present image processing technique to a high-compression PDF is given.
[Fourth embodiment]
The image forming apparatus of the fourth embodiment has the same configuration as the image forming apparatus shown in FIG. Accordingly, the same reference numerals as those of the image forming apparatus according to the first embodiment are used and a detailed description of the configuration is omitted. Note that the image forming apparatus in the present embodiment is an apparatus that detects and corrects an input tilt of a document.

Since a document image contains many characters, if a character pixel is projected onto the x-axis while rotating the image at a predetermined angle step, the inclination angle can be detected from the sharpness. However, if the rotation angle is processed in fine steps, the process of rotating the image takes a lot of time, so that the processing time becomes large, and if the rotation angle is processed in a rough step, a tradeoff occurs that accuracy decreases.
On the other hand, in the case of a document image including a large number of line segments such as a table, if the line segments can be extracted, the tilt angle can be detected at a relatively high speed only by obtaining the tilt angle of the line segment.

  Next, the operation of the image processing apparatus will be described with reference to FIG. The first identification unit 102 extracts and labels character / line segment pixels. Specifically, edge extraction and labeling of the edge pixels are performed by the same method as in the first embodiment. Furthermore, a circumscribed rectangle of the connected component is generated from the maximum and minimum values of the x-coordinate and y-coordinate of the pixels with the same label. If the circumscribed rectangle satisfies the following condition, the diagonal line of the circumscribed rectangle is regarded as a line segment.

Condition 1: The aspect ratio is greater than or equal to a certain threshold (for example, 100 or more).
Condition 2: (A) The upper right and lower left vertices of the rectangle are black pixels, and the upper left and lower right vertices are white pixels. Alternatively, (B) the upper left and lower right vertices are black pixels, and the upper right and lower left vertices are white pixels.

Next, the determination unit 103 determines whether or not skew (tilt) can be detected using the line segment information. Here, it is assumed that the line segments used in the document are horizontal or vertical.
The determination unit 103 obtains the average and variance of the inclination angles of the line segments identified by the first identification unit. The angle of each line segment is simply determined as θ = tan ⁻¹ (w / h). However, in the case of the pattern (B), the sign is reversed to θ = −tan ⁻¹ (w / h). Here, w and h are the width and height of the rectangle, respectively. In this way, the determination unit 103 calculates the inclination angle of each line segment. If the variance of the tilt angles is less than a certain threshold, it is determined that the average tilt angle θ of the line segments can be used as the skew angle of the entire document.

On the other hand, when the determination unit 103 determines that the inclination angle of the line cannot be used, the second identification unit 104 rotates the circumscribed rectangle regarded as a character by a predetermined angle step to the x axis. Projection is performed, and an inclination angle is detected from the histogram shape.
Here, since the character has a shape close to a square, an aspect ratio of the circumscribed rectangle that is less than a threshold (for example, 2) is regarded as a character. In addition, if the projection onto the x-axis is obtained while rotating the actual image, the processing time becomes enormous, so the projection is actually obtained by calculation.

FIG. 16 is a diagram for explaining a method for calculating a projected image. FIG. 16A shows the coordinates of a circumscribed rectangle in a state where there is no inclination. The upper left coordinates are (x0, y0), and the width and height are (sx, sy). FIG. 16B shows a state where the coordinate axis is rotated by an angle θ. The projection of the circumscribed rectangle onto the x-axis is in the range of α to β expressed by the following equations (2) and (3).
α = x0 cos θ−y0 sin θ Formula (2)
β = (x0 + sx) cosθ− (y0 + sy) sinθ (3)
This projection is performed for all circumscribed rectangles regarded as characters, and the angle θ at which the histogram shape of the projection length is the sharpest is detected as the skew angle of the document.

Next, an inclination correction method in the image processing unit 105 will be described. Assuming that the original inclination angle obtained by the first identification means or the second identification means is θ, in order to correct the inclination, the original has only to be rotated by the angle θ, so that the corrected pixel (x ′, The pixel value of y ′) is the pixel value of the pixel (x, y) that satisfies the following expression.

  Therefore, the pixel position (x, y) of the input image to be referred to obtain the pixel value at the corrected pixel position (x ′, y ′) is the inverse of the rotation matrix from the left side on both sides of the above equation (4). It is obtained by the following equation (5) by multiplying the matrix.

x = x'cosθ + y'sinθ
y = −x′sin θ + y ′ cos θ (5)
The coordinate position (x, y) of the input image obtained by Expression (5) often does not become an integer when the rotation angle θ is other than a multiple of 90 °. That is, the value often does not become a pixel unit. Therefore, using the coordinate position of pixel units (grid points) around the coordinate position (x, y) and its pixel value, for example, the input image pixel is rotated while being interpolated (bilinear method) as shown in Equation (6). To process. This prevents jaggy from occurring.

Note that (x ₀ , y ₀ ) and (x ₁ , y ₁ ) are the upper left lattice point and the lower right lattice point nearest to the coordinate position (x, y), and the upper left is the origin.

  As described above, in the present embodiment, when there is a reliable line segment in the document, the skew angle is detected using the line segment information, and only when it is determined that there is no reliable line segment. The skew angle is detected by the identification means. Thus, skew correction can be performed without wasting processing time.

[Fifth embodiment]
The image forming apparatus of the fifth embodiment has the same configuration as the image forming apparatus of the third embodiment shown in FIG. Accordingly, the same reference numerals as those of the image forming apparatus according to the third embodiment are used and the detailed description of the configuration is omitted.
The image forming apparatus according to the fifth embodiment generates the highly compressed PDF file described with reference to FIG. 3, and detects and corrects the orientation and input tilt of the document. In this embodiment, skew detection uses only character pixels without using line segments.

  Currently used skew detection and document orientation detection are processes based on the premise that many characters are included in an input document image. However, when the number of characters is small, the detection accuracy is low.

  Therefore, in the fifth embodiment, as in the third embodiment, the first identification unit 1002 (FIG. 10) performs edge extraction, and the determination unit 1003 executes a labeling process to calculate the number of circumscribed rectangles. Count. If the number of circumscribed rectangles is less than a predetermined threshold value (for example, 100), it is determined that sufficient skew / document orientation detection accuracy cannot be obtained, and the second identification unit 1004 is not executed. If it is determined that skew / document orientation detection is not performed, the first image processing unit 1005 does not perform skew / document orientation correction, and generates a highly compressed PDF by the method described in the first embodiment.

  When the determination unit 1003 determines that the skew / document orientation should be detected, the second identification unit 1004 performs the skew / document orientation detection process. The skew detection may be performed in the same manner as in the fourth embodiment, and the description is omitted. Document orientation detection is performed using OCR. First, an arbitrary character string is extracted, the image is rotated in four directions of 0 °, 90 °, 180 °, and 270 ° with respect to the character string image, and the OCR similarity is calculated. The angle with the highest similarity is detected as the correct document orientation. As described above, when the second identification unit detects the document tilt and the document orientation, the second image processing unit 1006 corrects the document tilt and the document orientation, and executes high-compression PDF creation.

  In the image forming apparatus according to each of the embodiments described above, the image area identification process is divided into the first identification process in the first half and the second identification process in the second half, and the second identification unit is implemented according to the input document image. Judge whether to do. Here, the first identification process is a process whose accuracy and speed do not depend on the input document image. In addition, the determination means for determining whether or not to implement the second identification means inputs or generates a signal to be used by the second identification means and makes a determination based on the signal.

  As a result of determining the complexity of the document, the second identification process is performed based on the first identification signal output from the first identification unit and the determination signal output from the determination unit when the document is a simple document. Then, image processing is performed based on the second identification signal. On the other hand, when it is determined that the document is complicated, the second identification process is not performed, and the image processing is performed using the first identification signal.

  Note that the functions of the first identification unit and the second identification unit are not classified based only on whether the accuracy / speed depends on the input document image. FIG. 17 is a diagram showing the functions of the first and second identification means classified. The first identification means and the second identification means of the present invention can be classified by the function shown in FIG. 17, and conversely, the first identification means and the second identification means classified by this function are It is included in the technical scope of the present invention.

  According to the image forming apparatus of these forms, when a complicated image is input, an image close to the user's desire is formed by reducing the risk of enormous processing time and image quality failure. It becomes possible. In addition, when a non-complex (standard) image is input, the first identification unit, the determination unit, and the second identification unit are configured only by processing that is originally necessary. The output image expected by the user can be obtained without affecting the identification accuracy.

Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage.
Further, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, you may combine suitably the component covering different embodiment.

  The present invention reduces an increase in image area identification processing time depending on an input document image, reduces an image quality deterioration risk of an output image, and can obtain an output image close to a user's expectation. Can be used in the manufacturing industry.

        DESCRIPTION OF SYMBOLS 101 ... Image input means, 102 ... First identification means, 103 ... Determination means, 104 ... Second identification means, 105 ... Image processing means, 301 ... Input image, 302 ... Background image, 302 ... Multi-value image, 303 ... character image, 303 ... character image, 401 ... image input means, 402 ... identification means, 403 ... edge extraction means, 404 ... outer rectangle generation means, 405 ... character string / character area extraction processing means, 406 ... image processing means 801 ... Image input means 802 ... First identification means 803 ... Pre-processing means 804 ... Determination means 805 ... Second identification means 806 ... Image processing means 1002 ... First identification means 1003 ... Judgment means, 1004... Second identification means, 1005... First image processing means, 1006.

JP-A-2005-175541 JP 2003-008909 A JP 2000-020726 A

Claims (10)

Image input means for acquiring image information from a document including a paper document and generating an input image;
First identification means for performing a first image area identification process on an input image and outputting a first identification signal indicating an attribute of the image;
Second identification means for inputting the first identification signal, performing a second image area identification process subsequent to the first image area identification process, and outputting a second identification signal indicating an attribute of the image;
Judgment means for inputting the first identification signal and outputting a judgment signal indicating whether or not to execute the second identification means;
When the determination signal indicates that it should not be executed, the first identification signal is selected as the third identification signal without performing the second image area identification processing, and the determination signal is executed. A selection means for selecting the second identification signal as the third identification signal;
An image forming apparatus comprising: an image processing unit that inputs the input image and the third identification signal and executes image processing without performing the second image area identification processing. Image input means for acquiring image information from a document including a paper document and generating an input image;
First identification means for performing a first image area identification process on an input image and outputting a first identification signal indicating an attribute of the image;
Pre-processing means for inputting the first identification signal, performing pre-processing of a second image area identification process following the first image area identification process, and outputting a pre-process signal;
Second identification means for inputting a pre-processing signal, performing the second image area identification process, and outputting a second identification signal indicating an attribute of the image;
Judgment means for inputting a pre-processing signal and outputting a judgment signal indicating whether or not to execute the second identification means;
When the determination signal indicates that it should not be executed, the first identification signal is selected as the third identification signal without performing the second image area identification processing, and the determination signal is executed. A selection means for selecting the second identification signal as the third identification signal;
An image forming apparatus comprising: an image processing unit that inputs the input image and the third identification signal and executes image processing without performing the second image area identification processing. Image input means for acquiring image information from a document including a paper document and generating an input image;
First identification means for performing a first image area identification process on an input image and outputting a first identification signal indicating an attribute of the image;
Second identification means for inputting the first identification signal, performing a second image area identification process subsequent to the first image area identification process, and outputting a second identification signal indicating an attribute of the image;
Judgment means for inputting the first identification signal and outputting a judgment signal indicating whether or not to execute the second identification means;
When the determination signal indicates that it should not be executed, the second image area identification processing is not performed, and the input image and the first identification signal are input to execute image processing. When the image processing means indicates that the determination signal should be executed, the image processing means includes second input image processing means for inputting the input image and the second identification signal to execute image processing. An image forming apparatus.   The first identifying means identifies a pixel attribute by referring only to the pixel and its vicinity, and the second identifying means identifies the attribute by referring to a wider area or the entire document. The image forming apparatus according to claim 1 or 2.   The first identifying means is an identifying means that does not depend on the distribution of pixel values of the input image, and the second identifying means is an identifying means that depends on the distribution of pixel values of the input image. The image forming apparatus according to claim 1, wherein the image forming apparatus is an image forming apparatus.   The first identification unit is an identification unit whose processing accuracy does not depend on the distribution of pixel values of the input image, and the second identification unit is an identification unit whose processing accuracy depends on the distribution of pixel values of the input image. The image forming apparatus according to claim 1.   4. The identification means according to claim 3, wherein the first identification means is an identification means whose processing time and accuracy do not depend on a distribution of pixel values of an input image composed of threshold processing of the pixel itself and filter processing of a fixed size. Item 5. The image forming apparatus according to Item 4.   The second identifying means is an identifying means whose processing time and accuracy depend on a distribution of pixel values of an input image for analyzing a mutual relationship between a size and a position of connected components of pixels classified by the first identifying means. The image forming apparatus according to claim 3, wherein the image forming apparatus is an image forming apparatus.   The first identifying means is means for identifying an attribute for each pixel of the input image, and the second identifying means is means for identifying an attribute for each area represented by a rectangle in the input image. The image forming apparatus according to claim 1. Obtain image information from a manuscript including a paper manuscript to generate an input image,
Performing a first image area identification process on the input image to generate a first identification signal indicating an attribute of the image;
Performing a second image area identification process subsequent to the first image area identification process on the first identification signal to generate a second identification signal indicating an attribute of the image;
Generating a determination signal indicating whether to execute the second identification means from the first identification signal;
When the determination signal indicates that it should not be executed, the first identification signal is selected as the third identification signal without performing the second image area identification processing, and the determination signal is executed. If it should indicate, the second identification signal is selected as the third identification signal,
An image processing method, wherein image processing is performed without performing the second image area identification processing on the input image and the third identification signal.

Images