JPS627590B2 - - Google Patents

Description

[Detailed Description of the Invention] The present invention is applicable to documents where line figures such as vertical lines or horizontal lines and characters intersect with each other, such as in slips, etc. Another object of the present invention is to provide a new image extraction method that can extract line figures and characters separately.

Note that hereinafter, characters, symbols, and numbers drawn on a document will be referred to as "characters" to distinguish them from other line figures such as frame lines.

Methods have been seen in the past to extract line figures and characters, or to extract only the characters, from the image information of a document on which line figures and characters are drawn, such as a slip, but in these cases, The condition for line figures and characters on a document is that ``line figures and characters do not intersect with each other.'' Using this condition, when the continuity of the black level signal is interrupted, the continuous area is Line figures and characters are distinguished by the size of the area they occupy, and these extraction methods cannot be applied to manuscripts in which there are parts where line figures and characters intersect. However, in practice, for example, when characters are written outside the frame lines of documents, etc., this frequently occurs, and this has become a very difficult problem in image extraction processing.

The present invention is an image extraction method that can extract line figures and characters separately even if there are parts where they intersect with each other on a document. I will explain based on this.

In FIG. 1, reference numeral 1 denotes a scanning section, which scans the document line-by-line and extracts image information on the document as a binary digital signal pixel by pixel.
The information is stored in the storage unit 2. Hereinafter, the direction of the line from which image information is extracted in one scan will be referred to as the main scanning direction, and the direction perpendicular to this will be referred to as the sub-scanning direction.

Reference numeral 2 denotes a first storage section, which stores the binary signals taken out by the scanning section 1 in scanning order, and has a capacity of one page or more of a document.

Reference numeral 3 denotes a black run detection section which checks the binary signal in the first storage section 2 to find black runs that may be part of vertical lines and horizontal lines.

Reference numeral 4 denotes a line extraction unit which checks the binary signals around the black runs that may be vertical and horizontal lines found by the black run detection unit 3 while leaving the information in a second storage unit 5, which will be described later. Then, vertical lines and horizontal lines are extracted, and information regarding this is stored in a third storage section 8, which will be described later.

A second storage section 5 stores information being processed by the line extraction section 4.

Reference numeral 6 denotes a memory correction unit, which stores image signals corresponding to the vertical lines and horizontal lines extracted by the line extraction unit 4 in the first storage unit 2.
At the same time, the first storage unit 2 is supplemented with image signals of the portions where these vertical lines and horizontal lines intersect with characters other than these.

Reference numeral 7 denotes a character extraction section, which extracts remaining characters, etc. from the image signal in the first storage section 2 after being processed by the memory correction section 6, and a third storage section 8 whose information related thereto will be described later. to be memorized.

Reference numeral 8 denotes a third storage section, which stores information regarding vertical lines and horizontal lines extracted by the line extraction section 4 and characters etc. extracted by the character extraction section 7.

FIG. 2 is an example of image information on a document used to explain this embodiment. Squares represent pixel divisions, a ₁ to a ₃₅ are addresses in the main scanning direction, and b _1
.about.b35 represents an address in the sub-scanning direction, but in order to avoid complicating the diagram, only every five addresses are given.

The processing method of this embodiment will be specifically explained below.

In this embodiment, the following restrictions are imposed on the nature of the image on the document. The thickness of the vertical line and horizontal line is in the range of 2 to 3 in terms of the number of pixels, and each column or row of pixels forming the vertical line or horizontal line has a length of 10 or more in terms of the number of pixels. Further, the lower limit of the length of the vertical line in the sub-scanning direction and the length of the horizontal line in the main-scanning direction, and the upper limit of the size of the character in the main-scanning direction and the sub-scanning direction are set to 20 in pixels. The above limitations have been set for the purpose of concretely explaining the processing, and do not limit the present invention.
This should be set based on the performance of the scanning unit, etc.

First, a document is line-sequentially scanned in the scanning section 1, and image information on the document is extracted pixel by pixel as a binary digital signal, which is stored for one page in the first storage section 2 in the scanning order.

Next, in the black run detection section 3, the first storage section 2
Examine the image signal within line by line in the main scanning direction, and check whether there are black runs with a length of 2 to 3, which corresponds to the thickness of the vertical line, and whether there are row elements with a length of 10 or more that form a horizontal line. Check. In the case of the manuscript example in Figure 2, the length of a ₂₈ to a ₂₉ in line b ₁ is 2.
The black run is detected as the first candidate for the vertical line, and this address information is taken over to the next line extraction unit 4.

The line extraction unit 4 sequentially traces the continuity of the vertical line candidate or horizontal line candidate detected by the black run detection unit 3 to the periphery of the black run for the lines before and after the line where the black run was detected. , it is determined whether or not the lines are within the limits regarding the vertical lines and horizontal lines described above, and lines that are within the limits are extracted as vertical lines or horizontal lines. In the case of the document example shown in FIG. 2, the black run detection unit 3 detects the runs a ₂₈ to a ₂₉ on line b ₁ as the first vertical line candidate, so the main scanning direction address a ₂₈ The pixel columns of and _a29 are checked in order from the sub-scanning direction address _b2 . Regarding pixel column a ₂₉ , black continues from b ₁ to b ₂₄ , satisfying the condition of a pixel column having a length of 10 or more and forming a vertical line. Also, the pixel column of a ₂₈ is b ₁
The black continues from b to _b15 , satisfying the condition of a pixel row of length 10 or more. Since the black of the pixel column a ₂₈ is interrupted at the point b ₁₅ , the pixel column a _{28 is}
When checking the pixel row of a ₃₀ on the opposite side, black continues from b ₁₃ to b ₂₄ and satisfies the condition of a pixel row of length _{10 or} more. It turns out that the conditions that the thickness is in the range of 2 to 3 and the minimum length of the vertical line in the sub-scanning direction are 20 or more are met, so this first vertical line candidate is determined to be a vertical line. The starting point address a ₂₈ , b ₁ and the ending point address a ₃₀ , b ₂₄ of this vertical line are stored in the third storage unit 8.
to be memorized. During this process, the second storage unit 5 is used, and when it is determined that it is a vertical line, it stores the starting and ending points of the three pixel columns forming this vertical line, a ₂₈ , b ₁ , b ₁₅ , a ₂₉ and
Only b ₁ , b ₂₄ , a ₃₀ and b ₁₃ , b ₂₄ are left.

In the memory correction unit 6, information regarding the element pixel column of the vertical line stored in the third memory unit 8 is read from the second memory unit 5, and the information related to the element pixel column of the vertical line stored in the third memory unit 8 is read out from the second memory unit 5, and
Erase the image signal within. After this, the image signal of the address corresponding to the left and right adjacent parts on the document of the element pixel row of the vertical line that has just been erased is checked, and if it is deemed to have intersected with the remaining image, it is deemed that it has intersected. The image signal of the part is set as the black signal. In the case of the manuscript example in Figure 2, the parts adjacent to the left and right of the vertical line are a ₂₇ and a ₃₀ between b ₁ and b ₁₂ , and b ₁₃
a ₂₇ and a ₃₁ between b ₁₅ and a ₂₈ between b ₁₆ and b ₂₄
and a ₃₁ pixel row. If there is black on the left side of the vertical line in this part, check whether there is black on the right side of the vertical line on three pixel rows including the previous and next pixel rows. If there is a black pair, the part with the vertical line between the black pairs is corrected to a black signal. In the manuscript example in Figure 2, b ₁
Looking at the pixel column _a27 between and _b12 , a black signal is detected in the pixel row _b8 , so this pixel row _{b8
The intersection point a 30 , b 7 ,} a ₃₀ ,
Checking b ₈ , a ₃₀ , and b ₉ reveals that a ₃₀ and b ₈ are black. This black signal pair a ₂₇ , b ₈ and a ₃₀ , b ₈
The points a ₂₈ , b ₈ , a ₂₉ , and b ₈ sandwiched between are corrected to black signals. As a result, the black signal in the portion corresponding to the intersection of the vertical line and the remaining image, which was once erased as a portion corresponding to the vertical line, is restored. The above is the processing of the vertical line starting from the black runs _a28 to _a29 on the _b1 line, and after this processing is completed, the processing returns to the black run detecting section 3 again.

Next, the black run processing unit 3 detects a black run with a length of 2 between a ₂₁ and a 22 on the b ₅ line as a black run with a length of _{2 to 3} corresponding to the thickness of the vertical line. Then, this is passed on to the vertical extraction unit 4 as a second vertical line candidate.

The line extraction unit 4 traces the pixel rows of the main scanning direction addresses a ₂₁ and a ₂₂ of this black run, but the pixel row of a ₂₁ starts at b ₅ and ends at b ₁₀ , and has a length of 4, a It is detected that _{the 22} pixel columns are pixel columns of length 5 starting at b ₅ and ending at b ₁₁ , and none of them are element pixel columns of the vertical line, which are pixel columns of length 10 or more. a ₂₁
a ₂₀ , a 19 , a 18 that appear on the left side of , a ₁₉ , a ₁₈ that appears on the right side of
The same applies to the pixel columns a ₂₃ , a ₂₄ , a ₂₅ , and a ₂₆ .

Next, extraction of horizontal lines will be explained.

The black run detecting unit 3 detects a black pixel row of length 14 starting from a ₁₈ and ending at a ₃₁ in line b ₈ , which satisfies the condition of a pixel row having a length of 10 or more and being an element of a horizontal line, so this is detected as the first black pixel row. The line extraction unit 4 takes over the line as a horizontal line candidate.

The line extraction unit 4 checks whether there is a black line with a length of 10 or more on the lines before and after the _b8 line where the first horizontal line candidate exists, but there are black lines on the _b7 line as well as on the _b9 line. Since there is no black run with a length of 10 or more, it is determined that this first horizontal line candidate is not a horizontal line part.

Next, in the black run detection section 3, on line b ₃₁
A black run of length 17 starting at a ₁₂ and ending at a ₂₈ is detected and handed over to the line extraction unit 4 as a second horizontal line candidate.

The line extractor 4 checks whether there are black runs with a length of 10 or more on the lines b ₃₀ and b ₃₂ before and after the line b ₃₁ containing the second horizontal line candidate. on line b ₃₂
A black run of length ₂₈ starting with a ₁ and ending with a 28 is detected, and a black run of length 28 starting with a ₁ and ending with a ₁₃ is detected on the next line b ₃₃ . From the above, line b ₃₁ ,
Each pixel row on b ₃₂ and b ₃₃ has a length of 10 or more, satisfies the thickness 2 to 3 between addresses a ₁ to a ₂₈ in the main scanning direction, and has an overall length of 28 is greater than or equal to the lower limit of the length of the horizontal line, 20, so this is determined to be a horizontal line, and its starting points a ₁ , b ₃₃ and ending points a ₂₈ , b ₃₁ are
The information is stored in the storage unit 8. In addition, as in the case of vertical line processing, after determining the second horizontal line candidate as a horizontal line, the second storage unit 5 stores the starting point and ending point of the three pixel rows forming the second horizontal line candidate, b ₃₁ and a ₁₂ , a ₂₈ , b ₃₂ and a ₁ , a ₂₈ , b ₃₃
and leave a ₁ and a ₁₃ .

The memory correction unit 6 reads information regarding the extracted element pixel row of the horizontal line from the second storage unit, and erases the corresponding image signal in the first storage unit 2. Thereafter, as in the case of vertical lines, the image signals of the portions adjacent above and below the horizontal line just erased are checked. In the case of Fig. 2, in the main scanning direction address range from a ₁ to a ₁₁ , the pixel rows are b ₃₁ and b ₃₄ , in the range from a ₁₂ to a ₁₃ , the pixel rows are b ₃₀ and b ₃₄ , and from a ₁₄ to pixel rows In the range of a ₂₈ , there are pixel rows of b ₃₀ and b ₃₃ . First a ₁₆ , b ₃₀
Since the dot is black, the adjacent main scanning direction address A ₁₆ on the opposite side with respect to the horizontal line and A ₁₅ and A ₁₇ before and after it are
Check the points, but there is no black signal. next
Since the points a ₁₇ , b ₃₀ are black, similarly a ₁₆ , b ₃₃ ,
When you check the three pixels a ₁₇ , b ₃₃ , a ₁₈ , b ₃₃ ,
Since we know that the pixels a ₁₈ and b ₃₃ are black, the four pixels sandwiched between these pixels a ₁₇ , b ₃₀ and a ₁₈ , b ₃₃ are a ₁₇ , b ₃₁ , a ₁₈ , b ₃₁ , a ₁₇ , b ₃₂ , a ₁₈ , b ₃₂ are corrected to black signals. As a result, the black signal in the portion corresponding to the intersection of the horizontal line and the remaining image, which was once erased as a portion corresponding to the horizontal line, is restored.

In this way, the vertical lines and horizontal lines on the document are extracted and their address information is stored in the second storage unit 5, the image signals corresponding to these vertical lines and horizontal lines are deleted from the first storage unit 2, and the vertical lines If there is a part where the horizontal line is considered to intersect with another image, the character extracting part 7 continues to extract the image signal in the first storage part 2 after being supplemented with the image signal of that part and converts it into a character. is extracted and the information is stored in the third storage unit 8.

According to the image extraction method according to the present invention, as shown in the above embodiment, even when there is a portion where a line figure and a character intersect on a document, restrictions regarding the thickness and length of vertical lines and horizontal lines can be applied. By setting, vertical lines, horizontal lines, and characters can be extracted individually using this restriction. Note that this limit can be set appropriately depending on the nature of the image on the document to be handled and the performance of the scanning unit, and provides an extremely effective method for actual image extraction processing.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the image extraction method according to the present invention, and FIG.
FIG. 6 is a diagram illustrating an example of image information on a document used to explain an example of processing. 1... Scanning section, 2... First storage section, 3... Black run detection section, 4... Line extraction section, 5... Second storage section, 6... Memory correction section, 7... Character extraction section, 8... Third storage section .

Claims (1)

[Claims] 1. Image information of a document on which line figures and characters are drawn is stored as a binary digital signal in pixel units, conditions are set regarding the thickness and length of the line figure, and images that satisfy the conditions are extracted as line figures. Then, a binary signal corresponding to image information of the line figure is deleted from the memory, and a binary signal corresponding to image information of a portion where the line figure and another image intersect is added to the memory. An image extraction method characterized by extracting characters from the stored image information after doing so.