JP3976804B2 - Image processing method - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to an image processing method and an image processing apparatus capable of correctly cutting out characters of a document image including a side line such as an underline or a surrounding frame in units of characters.
[0002]
[Prior art]
An example of recognition processing performed in the conventional OCR is shown in FIG.
[0003]
First, in S2601, a document image is input using an image scanner or the like.
[0004]
In step S2602, an image block is extracted from the input document image. An example of image block extraction using projection is shown in a horizontally written document. As shown in FIG. 27, by taking a projection 2701 in the left-right direction, extracting lines, and then taking a projection 2702 perpendicular to the line direction for each line. An image block 2603 can be extracted.
[0005]
In this state, there is a separation block 2705 in which one character is erroneously separated into two or more blocks, and there is a combination block 2706 in which two or more characters are erroneously combined and extracted as one block. If the reference character size is obtained from the average of the image block widths or the most frequent distribution of the image block widths, and if a plurality of image blocks are combined in step S2604, the reference character size is obtained. If the image block is an integral multiple of the reference character size, the image block is determined in character units by dividing it into integers. The extracted image block stores information related to the position, width, and height of the image block in step S2605.
[0006]
Thereafter, in S2606, each character image block is subjected to identification calculation, and the category (character) having the highest similarity is set as the recognition result. The result is stored in S2607.
[0007]
For vertical writing documents, the recognition processing is performed in the same manner by simply switching the horizontal and vertical directions.
[0008]
In this way, the OCR is configured to perform character segmentation after image input in order to perform identification calculation in units of characters.
[0009]
[Problems to be solved by the invention]
However, in the above conventional example, when there are image blocks including a side line, especially when a line and a character as shown in FIG. 5 are in contact with each other, they are extracted as one image block as in 501. There was a drawback that the character image could not be cut out correctly.
[0010]
When the character cutout method in the conventional character recognition includes a character surrounded by a frame as shown in FIG. 17, the frame character is extracted as one image block as shown in 1701, so that the image is displayed in character units. There was a drawback that could not be cut out correctly.
[0011]
[Means for Solving the Problems]
In order to solve the above problems, an image processing method according to the present invention includes an image block extraction step of extracting an image block including a character image from image blocks cut out by taking a projection from an original image, and the image For the image block extracted in the block extraction step, it is determined whether there is a possibility that a line image is included based on the width of each image block and the reference character size, and the line image may be included. A determination step that sets an image block determined to be a target image block as a target image block, a tilt calculation step that determines a tilt of a character image in the target image block determined in the determination step, and the tilt in the target image block A histogram calculation step for taking a histogram in a direction along the inclination obtained in the calculation step; and the histogram calculation step The line image information extraction step for extracting line image information based on the position and width of the peak whose histogram frequency calculated in step 1 is equal to or greater than a predetermined threshold, and the line extracted by the line image information extraction step An extraction step of extracting image information obtained by removing the line image from the target image block based on image information.
[0012]
In order to solve the above problems, an image processing apparatus of the present invention includes an image block extracting unit that extracts an image block including a character image from image blocks cut out by taking a projection from an original image, and the image For the image block extracted by the block extraction means, it is determined whether there is a possibility that a line image is included based on the width of each image block and the reference character size, and the line image may be included. A determination unit that determines an image block that has been determined to be a target image block; a tilt calculation unit that determines a tilt of a character image in the target image block determined by the determination unit; and the tilt in the target image block A histogram calculating means for taking a histogram in a direction along the inclination obtained by the calculating means; and a frequency of the histogram calculated by the histogram calculating means. Line image information extraction means for extracting line image information based on the position and width of the peak that is equal to or greater than a predetermined threshold, and based on the line image information extracted by the line image information extraction means, Extracting means for extracting image information obtained by removing the line image from the target image block.
[0029]
【Example】
Example 1
FIG. 1 is a block diagram showing the configuration of a character recognition apparatus for carrying out the present invention. 101 is a CPU that controls the entire apparatus in accordance with a control program stored in the ROM 102, 102 is a ROM that stores the control program of the apparatus, such as the processing shown in the flowchart described below, executed by the CPU 101, and 103 is a document image. A RAM to be stored, 104 is a display, and 105 is a keyboard. Various commands and images and text displayed on the display 104 are edited from the keyboard 105. Reference numeral 106 denotes an image scanner for optically reading an original document image.
[0030]
An outline of this embodiment executed by the character recognition apparatus having the configuration shown in FIG. 1 will be described with reference to the flowchart of FIG.
[0031]
First, in step S <b> 201, a document image of a document placed on the image scanner 106 is input and stored in the RAM 103. The input image stored in the RAM 103 may be displayed on the display 104. When a recognition frame for performing recognition processing is set on the display 104, the recognition frame information is also stored in the RAM 103. The image input is not limited to the image scanner, but may be read from a storage device (not shown) or may be transmitted from another computer by communication means.
[0032]
In step S202, image blocks are extracted from the input document image stored in the RAM 103 in the same manner as in step S1402 in the conventional example. In step S203, the reference character size is obtained in the same manner as in step S1403. The information obtained in S202 and S203 is stored in the RAM 103.
[0033]
Here, for each image block extracted in S202, it is sequentially determined for each block whether or not it is a block including a side line in S204. If the target block is a block including a side line in S205, an image obtained by removing the side line from the block is extracted in S206, and a new image block including no side line is obtained.
[0034]
When all the image blocks have been extracted, in S207, the image blocks obtained in S203 and extracted using the reference character size stored in the RAM 103 are further divided to determine character-by-character image blocks. In S208, the extracted image block information (reference coordinates, width, height) in character units is stored and stored in the RAM 103 together with position information in the input image and information on appearance order (output order) in character units. In S209, the stored image block information is read one by one, and each character image block is identified with the dictionary data of the character pattern stored in the ROM 102, and the category (character) having the highest similarity is recognized. As a result. The result is stored in the RAM 103 together with position information of the character image obtained as a result in S210 in S210. The recognition result may be displayed on the display 104. In S210, in addition to the character having the highest similarity, a plurality of upper characters may be stored as candidate characters.
[0035]
The determination of the block including the side line in S204 will be described in detail using the flowchart of FIG. In S301, a ratio R = (width of target image block) / (reference character size) with respect to the reference character size of the width of the target image block is obtained. In S302, a predetermined threshold value T and R are compared, and if R is greater than T, the process proceeds to S303, and the block includes a side line. If R is smaller than T, the process proceeds to S304, and the block does not include a side line. Is determined. Usually, a side line is attached over a plurality of characters, so T is set to 2 or more. Further, in the case where there is a side line in the entire line and the reference character size cannot be obtained from the average of the image block widths or the distribution of the image block widths, the reference character size may be set to the line height in S203. In S204, the processing shown in the flowchart of FIG. 3 is performed on all the image blocks extracted in S202, and identification information as to whether or not all blocks are side line blocks is added.
[0036]
In S205, an image block to which identification information indicating that it is a sideline block added in S204 is added is determined, and the process proceeds to a process of removing a sideline from the image block in S206.
[0037]
The extraction of the image block excluding the side line in S206 will be described in detail using the flowchart of FIG. First, in step S401, information on the side line necessary for extracting the image block is extracted from the image block including the side line of interest. In the present embodiment, as the side line information, the width of the side line, the inclination of the side line, and the position of the side line are obtained.
[0038]
As shown in FIG. 6, the slope θ of the side line is set to two regions 602 at an appropriate interval L in the target image block, and a projection 601 in the region is obtained. From the projection deviations d and L, tan θ = The slope θ can be obtained as d / L. The projection to be obtained at this time is a method of searching for pixels in the left-right direction and setting it to 1 if there is even one black pixel. The interval L may be set to a position that is moderately inside (for example, a 1/2 character width) from the width depending on the width of the target image block. Alternatively, if the block width is sufficiently large (for example, R = 10), an appropriately determined L (for example, twice the reference character size) may be determined at an appropriate position.
[0039]
The position of the side line is below the character (underline), above (overline) or both above and below. In order to obtain the position of the side line, as shown in FIG. 7, the pixel block of interest 701 tilted by the angle θ (here, an image including two characters of “side line” and an underline) is 1 for each 1 / tan θ pixel. Taking a histogram of black pixels while shifting up and down pixel by pixel. The result is shown in FIG. In many cases, the histogram exists, for example, from under the image block to ¼ of the height if it is underlined, to ¼ of the height from the top of the image block if it is overlined. If the maximum value of the histogram is sufficiently larger than the width of the image block (the threshold value H determined by the block width) _TH If it is larger), it is determined that there is a side line in the portion, and as a result, H of the histogram _TH If the maximum value (or the next peak) exceeding ¼ is ¼ from the top, the overline can be obtained, and if it is ¼ from the bottom, the underline and the side line can be obtained. In addition to the method of taking a histogram, the continuity of black pixels is observed while shifting one pixel at a time for each 1 / tan θ pixel, and there is a black run sufficiently long with respect to the width of the image block. Even if a run is cut off halfway, if the width of the cut portion is very narrow, the cut portion is due to printing accuracy, and it is determined that the portion is connected, that is the sideline image block There is also a method for determining a long enough line as a side line.
[0040]
As shown in FIG. 8, the width W of the side line is a black pixel histogram of the part where the side line taken while shifting by 1 pixel for each 1 / tan θ pixel, and for the part including the outermost peak, Predetermined value H _TH It is obtained by the width w of the portion where the frequency is larger (for example, half the block width). If this width is smaller, the width of the side line may be w + α.
[0041]
As a result, if sideline information is not obtained, the image block extraction process ends in S402.
[0042]
If the information on the side line is obtained, the image block is extracted so as not to include the side line in S403. When extracting an image block in units of characters using the vertical projection as described in the conventional example, as shown in FIG. 9, while excluding the side line width w, and shifting by 1 pixel per 1 / tan θ pixel. The projection 901 can be extracted by taking the projection, and as a result, an image block 902 in character units in which “side” and “line” are divided is obtained. In S206, the processing shown in the flowchart of FIG. 4 is sequentially repeated for all the blocks determined as the side line blocks in S205.
[0043]
As described above, by extracting the information of the side line for the block determined to include the side line and extracting the image block only for the information obtained, the determination of the side line block is performed. Even if the accuracy is not high, there is an effect of correctly extracting a character image from an image block including a side line.
[0044]
Further, by obtaining the inclination of the side line, there is an effect of accurately extracting the character image even for the image block having the inclination.
[0045]
In this embodiment, the description is made with the minimum configuration requirements for carrying out the present invention. However, for example, the operator can specify a recognition area or perform an operation of correcting the recognition result. There is no problem even if the recognition area is automatically determined without intervention of an operator, or a process for reducing erroneous recognition is added.
[0046]
In this embodiment, a horizontally written document has been described. However, vertical writing can be performed in the same manner by only changing the row direction and the column direction.
[0047]
Further, the general-purpose computer may be configured to externally provide a program for performing processing for carrying out the present invention and store the control program for the apparatus in the RAM, and the storage device is limited to the RAM. not.
[0048]
In this embodiment, the example in which the processes of S204 and S206 are repeated for all the blocks has been described. However, after the processes of S204 to S206 are performed for one block, the processes of S204 to S206 are performed for the next block. You may set a loop to apply.
[0049]
(Example 2)
In the flowchart of FIG. 2, another example of the determination of the block including the side line in S204 and the extraction of the image line in the side line in S206 will be described with reference to FIGS. 10 and 11. Others are the same as those in the first embodiment, and a description thereof will be omitted.
[0050]
The determination of the block including the side line in S204 will be described using the flowchart of FIG. In S1001, the upper projection is taken. For example, a projection from the top of the image block to ¼ of the height may be taken. In S1002, the maximum width L1 of the upper projection frequency is obtained. The lower projection is taken in S1003, and the maximum width L2 of the lower projection frequency is obtained in S1004. If there is a side line, the projection width becomes a value close to the width of the image block. If the threshold obtained from the width of the image block is L, it is determined in S1005 that a side line is included if L1> L or L2> L. If not, it is determined that no side line is included, and the process proceeds to S1010. If the width of the image block is W, L = 0.9 × W may be set. If it is determined that a side line is included, information about the side line is extracted in S1006. The extraction method may be the same as in S401. If the information of the side line is not extracted, the process proceeds to S1010, and the image block is an image block that does not include the side line. In step S1008, the extracted sideline information is stored and stored in the RAM 103, and in step S10009, the target image block is changed to a block including a sideline, and the process ends. Note that the order of obtaining L1 and L2 may be reversed. Further, as a method of extracting the information of the side line in S1006, a point corresponding to a line with a width of L is searched for the image block from the bottom if it is an underline, or from the top if it is an overline. By calculating d from the difference and deriving θ, faster processing is realized.
[0051]
Another method of extracting the image block for the side line in S206 will be described with reference to the flowchart of FIG. In step S <b> 1101, image blocks are extracted so as not to include a side line using the side line information stored in the RAM 103. The extraction method may be the same as in S403.
[0052]
As described above, the local projection of the image block is used for the determination of the block including the side line, and the width thereof is used, so that the determination accuracy is improved. Further, by extracting the information of the side line at that time, there is an effect that the processing of extracting the image block is performed on the image block that does not include the side line and the processing time is not wasted.
[0053]
(Example 3)
The outline of the processing of another embodiment executed by the character recognition apparatus having the configuration shown in FIG. 1 will be described with reference to the flowchart of FIG.
[0054]
In step S1201, a document image is input. In step S1202, an image block is extracted from the input document image. In step S1203, a reference character size is obtained. In step S1204, it is determined whether the block includes a side line. If the target block is a block including a side line in S1205, an image block excluding the side line is extracted in S1206. The steps so far are the same as those in the first embodiment and have been described in detail in S201 to S206, and thus detailed description thereof is omitted here. If the image block obtained in S1206 still contains a side line, it is checked again in S1211 whether or not to extract the image block excluding the side line, and if so, S1206 is executed again. The determination method in S1211 may be the same as that in S1204, and the image block may be extracted again even when the horizontal width of the image block does not change before and after the side line extraction.
[0055]
When all the image blocks have been extracted, an image block in character units is determined using the reference character size in S1207, the image block information extracted in S1208 is stored, an identification operation is performed in S1209, and the result is obtained in S1210. Remember. The processing of S1207 to S1210 is the same as that of S207 to S210.
[0056]
As described above, after extracting the image block from the image block including the side line and extracting the image block again, the double or multiple side line is determined by determining whether or not to extract the image block excluding the side line again. It is also possible to extract a character image block even for an image block including an image block or a partially multiplexed image block.
[0057]
Example 4
Storage of the calculation result of S210 in the flowchart of FIG. 2 will be described with reference to FIG. If the result of the identification operation in FIG. 5 is “This is a side line”, an underline start control code UI1301 is placed before “side”, and an underline end control code UO1302 is placed after “line” in FIG. The data is stored in the RAM 103 as shown in a). If a side line can be displayed on the display 104, an underline is drawn from the UI to the UO character to display the recognition result. Similarly, control codes may be determined for overlines and multiple lines.
[0058]
If each character has an attribute related to a side line instead of a control code, the attribute of “side” and “line” may be underlined, and the result may be stored as shown in FIG.
[0059]
Note that the data structure of the result is not limited to that shown in FIG. 13, and information such as candidate characters may be included therein.
[0060]
As described above, by storing characters with a side line, there is an effect that information other than characters of the input image can be reproduced without loss.
[0061]
(Example 5)
In the above-described first embodiment, the method of correctly cutting out character units from image information including a character string with an underline such as an underline as shown in FIG. 5 has been described. In this embodiment, as shown in FIG. A method for cutting out characters correctly from an image including a character string surrounded by a frame will be described.
[0062]
The present embodiment is implemented by the character recognition device having the configuration described in detail with reference to FIG. 1 in the first embodiment.
[0063]
An outline of this embodiment executed by the character recognition apparatus having the configuration shown in FIG. 1 will be described with reference to the flowchart of FIG. Here, the characteristic processing of this embodiment is from the processing for finding a surrounding frame block in S1401 to the processing for removing the surrounding frame in S1403. Here, these three steps will be described in detail. The same processing steps as those shown in the flowchart of FIG. 2 described in the first embodiment are denoted by the same step numbers. A description of this process is omitted here.
[0064]
For each image block extracted in S203, whether or not it is a block including an enclosing frame is sequentially determined for each block in S1401. If the target block is a block including an enclosing frame in S1402, an image obtained by removing the enclosing frame from the image of the block is extracted in S1403 to obtain a new image block that does not include the enclosing frame.
[0065]
The determination of the block including the enclosing frame in S1401 will be described in detail with reference to the flowchart of FIG. In step S1501, a ratio R = (width of target image block) / (reference character size) with respect to the reference character size of the width of the target image block is obtained. In S1502, a preset threshold value T is compared with R, and if R is smaller than T, the process proceeds to S1503, and the frame is determined to be a block that does not include a surrounding frame. If R is larger than T, the process proceeds to S1504 and proceeds to a process of confirming whether the frame is a surrounding frame block. In S1504, the projection of the upper part of the image block is taken. For example, a projection from the top of the image block to ¼ of the height may be taken. In S1505, the maximum width L1 of the upper projection taken in S1504 is obtained. If there is a surrounding frame, the projection width becomes a value close to the width of the image block. In S1506, the threshold obtained from the width of the image block is set to L (for example, 0.9 times the block width), and L and L1 are compared. . If L1 <L, it is determined that the block is not a surrounding frame, and the process advances to S1503. Similarly, the lower projection is taken in S1507, and the maximum width L2 of the lower projection is obtained in S1508. If L2 <L in S1509, it is determined that the frame is not a frame, and the process proceeds to S1503. Otherwise, the flow proceeds to S1510, and the frame is determined to be a frame. In S1503 and S1510, identification information that can be used to determine whether or not the image block is a frame block is added to the image block of interest and stored in the RAM 103. Note that the order of obtaining L1 and L2 may be reversed.
[0066]
The extraction of the image block for the surrounding frame in S1403 will be described in detail with reference to the flowchart of FIG. First, in step S <b> 1601, information on a surrounding frame necessary for extracting an image block is extracted from the image block of interest. In this embodiment, the line width w of the surrounding frame and the inclination θ of the surrounding frame are obtained.
[0067]
As shown in FIG. 18, the inclination θ of the surrounding frame is set to two regions 1800 with an appropriate interval L (the setting method of L is the same as in the first embodiment) in the target image block, and the image data of each region A projection 1801 is obtained for, and θ is obtained as tan θ = d / L from the deviations d and L of the projection. If it is obtained accurately, the inclination is calculated above and below the projection. If the values are close, the average is the inclination of the target image block. If the values are not close, the inclination is not determined, and the inclination value is not determined. Since the frame information cannot be extracted, the processing of the flowchart of FIG. 16 for the image block is ended.
[0068]
As shown in FIG. 19, the line width w of the enclosing frame is obtained by taking a histogram of black pixels while shifting the pixel block of interest up and down by 1 pixel for every 1 / tan θ pixel, and by placing the peak at the outermost side of the image block. For the part including _TH Obtained by the width w of the larger portion. If this width is smaller, the line width of the surrounding frame may be w + α. If it differs between the upper and lower sides of the enclosing frame, both the upper and lower line widths of the frame are obtained.
[0069]
If the information on the surrounding frame is not obtained at this stage, it is determined that the information on the surrounding frame cannot be extracted in S1602, and the image block extraction processing in the flowchart of FIG. The frame tilt θ and the frame width w extracted in S1601 are stored in the RAM 103 in association with the image block.
[0070]
If the information about the surrounding frame is obtained, image blocks are extracted so as not to include the surrounding frame in S1603. When extracting an image block in units of characters using the projection in the vertical direction as described in the conventional example, as shown in FIG. 20, the line width w of the surrounding frame is not included in the vertical direction, and every 1 / tan θ pixel. The projection 2001 can be extracted by taking the projection while shifting the pixel by one pixel, and as a result, an image block 2002 is obtained. Of the obtained blocks, the left and right end blocks 2003 are part of the enclosing frame as long as the width is close to the line width w of the enclosing frame. Therefore, the left and right end blocks are not character image blocks. If there is no block having a width approximately equal to the line width of the surrounding frame on the left and right of the target image block, it is determined that the image block does not include the surrounding frame.
[0071]
(Example 6)
Another embodiment of determining a block including a surrounding frame in S1401 and extracting an image block of the surrounding frame in S1403 will be described. Other processing steps are the same as those in the first embodiment, and a description thereof will be omitted.
[0072]
The determination of the block including the enclosing frame in S1401 will be described with reference to the flowchart of FIG. In S2101, the ratio R of the width of the image block of interest to the reference character size is obtained. If R is smaller than T in S2102, the process proceeds to S2103 to determine that the block does not include a surrounding frame. If R is larger than T, the outermost boundary line of the target image block is traced in step S2104. If the boundary line exceeds the permissible range, it is determined in S2104 that it is not a frame, and the process proceeds to S2103. An example of the permissible range will be described with reference to FIG. 22. In the left half of the block, when the y coordinate of the boundary line exists between y1 and y2 that is predetermined, the x coordinate is present to the right of the predetermined x1. For example, it is assumed that the boundary line does not track the bounding frame and exceeds the allowable range. If the x coordinate is between x1 and x2 in the upper half of the block, the allowable range is exceeded if the y coordinate is below y1. The same applies to the right half and the lower half.
[0073]
If the boundary line does not exceed the allowable range, the process advances to step S2106 to extract information on the surrounding frame. The extraction method may be the same as S1601 in the fifth embodiment. If the information of the surrounding frame is not extracted, the process proceeds to S2103 in S2107 and the determination ends. The extracted box information is stored and stored in the RAM 103 in S2108, and the process advances to S2109 to determine that the box is a box.
[0074]
The extraction of the image block of the enclosing frame in S1403 will be described with reference to the flowchart of FIG. Using the frame information stored in S2301, image blocks are extracted so as not to include the frame. The extraction method may be the same as S1603 in the first embodiment.
[0075]
As described above, the use of the boundary line for the determination of the block of the surrounding frame has an effect of improving the determination accuracy. Further, since the information of the surrounding frame is extracted by the determination of the block of the surrounding frame, there is an effect of preventing the image block from being extracted from the image block that is not the surrounding frame.
[0076]
(Example 7)
An outline of another embodiment different from the fifth embodiment of character segmentation excluding the enclosing frame executed by the character recognition apparatus having the configuration shown in FIG. 1 will be described with reference to the flowchart of FIG. However, the same processing steps as those shown in the flowchart of FIG. 14 of the fifth embodiment are denoted by the same step numbers, and description thereof is omitted here.
[0077]
A characteristic processing step in the present embodiment is S2401, and if the image block obtained in S1403 still includes an enclosing frame, it is checked in S2401 whether or not the image block excluding the enclosing frame is extracted again. If so, S1403 is executed again. This redetermination method may be the same as that in S1401.
[0078]
When all image blocks have been extracted, an image block in character units is determined using the reference character size in S207, the image block information extracted in S208 is stored, an identification operation is performed in S1209, and the result is obtained in S210. Remember.
[0079]
As described above, after extracting the image block from the image block including the enclosing frame and excluding the enclosing frame, it is determined whether the image block excluding the enclosing frame is extracted again. There is also an effect that a character image block can be extracted even for an image block including multiple surrounding frames.
[0080]
(Example 8)
Storage of the calculation result in S1410 in the flowchart of FIG. 14 will be described with reference to FIG. Assuming that the result of the identification calculation in FIG. 17 is “this is a box”, as shown in FIG. 25A, the control code FI2501 for starting the box before “box”, and the box behind the “frame”. An end control code FO 2502 is entered and stored in the RAM 103. If an enclosing frame can be displayed on the display 104, the recognition result is displayed by enclosing the encircling characters from FI to FO.
[0081]
In addition, if each character has an attribute related to an enclosing frame instead of a control code, the result is stored with each attribute of “enclosed”, “mi”, and “frame” as an enclosing frame as shown in FIG. Also good.
[0082]
Note that the data structure of the result is not limited to that shown in FIG. 13, and information such as candidate characters may be included therein.
[0083]
As described above, by storing characters with a surrounding frame, there is an effect that information other than characters of the input image can be reproduced without loss.
[0084]
【The invention's effect】
As described above, according to the present invention, since the image from which the line image is removed is extracted from the image information including the line image other than characters, it takes time and effort to erase the line on the document and input the image. In addition, it is possible to correctly cut out characters with lines added in character units.
[0085]
As described above, according to the present invention, an image can be correctly cut out in units of characters even when an image of a character to which a side line such as an underline or overline is added is input.
[0086]
As described above, according to the present invention, even when an image of a character surrounded by a surrounding frame is input, the image can be correctly cut out in units of characters.
[0087]
As described above, according to the present invention, whether or not the block includes a line image other than characters is determined based on the ratio between the width of the extracted image block and the reference character width. An image block to which a line is added can be determined.
[0089]
As described above, according to the present invention, the image information from which the line image is removed is extracted, so that it can be divided into character units with easy processing.
[0090]
As described above, according to the present invention, the information related to the line width is extracted, and the line image is removed according to this information. Therefore, even if line images having various line widths are added to the character, Line images can be removed.
[0091]
As described above, according to the present invention, information related to the inclination of the line is extracted, and line images are removed according to this information. Therefore, even when the document is inclined or the character string is inclined accurately. Line images can be removed.
[0092]
As described above, according to the present invention, the information related to the position of the line is extracted, and the line image is removed according to this information. Can be removed.
[0093]
As described above, according to the present invention, since it is determined again whether or not the image from which the line image is removed includes the line image, it is possible to completely remove the line image even if the characters have multiple line images added thereto. .
[0094]
As described above, according to the present invention, character recognition is performed after the line image is removed, so that the accuracy of character recognition is improved.
[0095]
As described above, according to the present invention, the line image attribute is added to the character even after the character is recognized, so that the line information originally added to the document is reflected when displaying the text. Can do.
[Brief description of the drawings]
FIG. 1 is a block diagram showing the configuration of a character recognition apparatus according to an embodiment of the present invention.
FIG. 2 is a flowchart of character recognition processing in the first embodiment.
FIG. 3 is a flowchart of detailed processing of side line block determination according to the first embodiment.
FIG. 4 is a flowchart of detailed processing of image extraction excluding side lines in the first embodiment.
FIG. 5 is a diagram illustrating an example of extracting an image block of an image including a character string to which a side line is added.
FIG. 6 is a diagram for explaining a method for detecting the inclination of a side line.
FIG. 7 is a diagram for explaining how to take a histogram when detecting the position of a side line
FIG. 8 is a diagram illustrating a method for detecting the width of a side line
FIG. 9 is an explanatory diagram of image extraction excluding side lines and image segmentation in character units.
FIG. 10 is a flowchart of detailed processing of side line block determination according to the second embodiment.
FIG. 11 is a flowchart of detailed processing of image extraction excluding side lines in the second embodiment.
FIG. 12 is a flowchart of character recognition processing according to the third embodiment.
FIG. 13 is a diagram showing a storage example of a character string to which an attribute of a side line is added
FIG. 14 is a flowchart of character recognition processing in the fifth embodiment.
FIG. 15 is a flowchart of detailed processing for enclosing frame block determination according to the fifth embodiment.
FIG. 16 is a flowchart of detailed processing of image extraction excluding a surrounding frame in the fifth embodiment.
FIG. 17 is a diagram illustrating an example of extracting an image block of an image including a character string to which a surrounding frame is added.
FIG. 18 is a diagram for explaining a method for detecting the inclination of a surrounding frame.
FIG. 19 is a diagram for explaining how to take a histogram when detecting the line width of an enclosing frame;
FIG. 20 is a diagram for explaining a method for detecting the width of a surrounding frame.
FIG. 21 is an explanatory diagram of extracting an image excluding a surrounding frame and extracting an image in character units.
FIG. 22 is a flowchart of detailed processing for enclosing frame block determination according to the sixth embodiment.
FIG. 23 is a flowchart of detailed processing of image extraction excluding a surrounding frame in the sixth embodiment.
FIG. 24 is a flowchart of character recognition processing according to the seventh embodiment.
FIG. 25 is a diagram showing a storage example of a character string to which an attribute of a frame is added
FIG. 26 is a flowchart of conventional character recognition processing.
FIG. 27 is a diagram showing an example in which a character image is divided by a histogram from image information where lines and characters are separated from each other;

Claims (12)

An image block extraction step for extracting an image block including a character image from among image blocks cut out by taking a projection from the original image;
For the image block extracted in the image block extraction step, it is determined whether there is a possibility that a line image is included based on the width of each image block and the reference character size, and the line image is included. A step of determining an image block that is determined to be a target image block,
An inclination calculating step for determining an inclination of the character image in the target image block determined in the determining step;
In the target image block, a histogram calculation step that takes a histogram in a direction along the inclination obtained in the inclination calculation step;
A line image information extraction step for extracting line image information based on the position and width of a peak at which the frequency of the histogram calculated in the histogram calculation step is equal to or greater than a predetermined threshold ;
An image processing method comprising: an extraction step of extracting image information obtained by removing the line image from the target image block based on information of the line image extracted by the line image information extraction step. Assuming that the inclination angle obtained in the inclination calculation step is θ, in the histogram calculation step, when taking a histogram in the direction along the inclination in the target image block, the position where the histogram is taken is one pixel in the vertical direction for each tan θ pixel. the image processing method according to claim 1, characterized in that go I taken a histogram of black pixels in the horizontal direction while shifting.   The image processing method according to claim 1, wherein the line image includes at least one of a side line and a frame of a surrounding frame.   The image processing method according to claim 1, wherein the line image includes at least one of an underline and an overline. The image processing method according to claim 1, further comprising a dividing step of dividing the image information extracted in the extracting step into character units. In the linear image information extraction step, as the information of the line image, the width of the該線image, gradient, and extracts information relating to the position,
2. The image processing method according to claim 1, wherein in the extraction step, image information obtained by removing the line image from the image is extracted based on information on the width, inclination, and position of the acquired line image. . 2. The image processing method according to claim 1, wherein the processing by the histogram calculation step, the line image information extraction step, and the extraction step is performed again on the image information extracted in the extraction step. .   2. The image processing method according to claim 1, further comprising a character recognition processing step of performing character recognition processing on the image information extracted in the extraction step to obtain a character code as a recognition result. Further, in the character recognition step, the recognition result obtained by performing character recognition processing on the image information obtained by removing the line image from the image determined to include the line image in the extraction step, The image processing method according to claim 8, wherein an attribute related to the line information is added and stored.   9. The image processing method according to claim 8, wherein in the character recognition step, a code related to the line information is inserted and stored in a character code string of a recognition result obtained by performing character recognition processing.   In the inclination calculating step, two small areas are set with a predetermined interval in the image, a projection is obtained in each small area, and based on the deviation between the projections of each small area and the predetermined interval The image processing method according to claim 1, wherein an inclination is obtained. Image block extraction means for extracting an image block including a character image from image blocks cut out by taking a projection from the original image;
For the image block extracted by the image block extraction means, it is determined whether there is a possibility that a line image is included based on the width of each image block and the reference character size, and the line image is included. A determination unit that sets an image block that has been determined to be a target image block,
An inclination calculating means for obtaining an inclination of the character image in the target image block determined by the determining means;
In the target image block, a histogram calculation unit that takes a histogram in a direction along the inclination obtained by the inclination calculation unit;
Line image information extraction means for extracting line image information based on the position and width of a peak whose histogram frequency calculated by the histogram calculation means is equal to or greater than a predetermined threshold ;
An image processing apparatus comprising: extraction means for extracting image information obtained by removing the line image from the target image block based on the line image information extracted by the line image information extraction means.

Images