JP3512290B2 - Image processing device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for detecting characters with high accuracy by using two types of edge detection methods. 2. Description of the Related Art There are two methods for detecting a character area. One of them is a method of detecting the magnitude of the inclination of an image, which is often a method of detecting a density difference between pixels relatively close to each other (for example, JP-A-7-30752 and JP-A-7-30752). No. 15607). Although this method can reliably detect an edge portion, it has a disadvantage that an edge portion other than a white background, that is, a contour portion in a picture region such as a photograph is erroneously determined as a character. Another method is to detect characters based on the density difference and the maximum and minimum densities (for example, Japanese Patent Publication No.
No. 28375) is superior to the first method in that a contour portion in a picture is not detected as a character, but when a small character has a low internal contrast,
There is a disadvantage that the inside of a character is not detected as a character. A method for solving this problem is disclosed in Japanese Patent Laid-Open Publication No. Hei 5-30.
No. 8516. This is a combination of a method using a parameter for detecting a character of normal density and a method using a second parameter for detecting the inside of a character. That is, the threshold values th1 and t
Edge detection based on a combination of h2 and edge detection based on a combination of thresholds th3 and th4 having small differences are performed.
By limiting the edge detected by the combination of the thresholds th3 and th4 to a range of a certain number of pixels following the pixel detected as the edge by the combination of the thresholds th1 and th2, the edge of the picture (photograph or halftone dot) is mistaken. While preventing detection, line image edges such as edges inside characters with low contrast are also detected with high accuracy. In recent years, it has become possible for a copying machine to automatically discriminate between a character and a picture and to perform an optimum process on each of them to perform high-quality copying. Such copying includes not only so-called ordinary printed documents and photographic paper photographs, but also copied documents output from a copying machine. [0006] By the way, toner is liable to be scattered at the time of output of the copying machine, and small characters or crowded characters tend to be crushed. In addition, the degree of scattering of the toner varies depending on the density, thickness, size, and the like of the character, and thus the character is crushed in various ways. When such a document is read by a scanner, the image density also becomes various levels. The method described in Japanese Patent Laid-Open No. 5-308516 cannot detect such characters having various density levels. FIGS. 7A and 7B are diagrams for explaining the detection inside a character. FIG. 7A shows a case where the inside of a character can be detected.
Indicates a case where detection is not possible. Using a combination of thresholds th1 and th2, it is determined that density ≦ th1 is white, and density ≧ th
If 2 is determined to be black, in the case of (a), both the outer edge and the inner edge of the character can be detected as a pair of white pixels and black pixels. However, in the case of (b), since the inside of the character is lower in contrast than the outside, a white pixel cannot be detected. That is, as shown in FIG. 7B, when the density inside the character is high, the edge cannot be determined as a character. However, even in such a case, since the density difference between pixels is large to some extent, it is possible to detect a character by detecting this density difference. SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and an object of the present invention is to detect a character based on the absolute density of a pixel outside a character, and to detect a difference in the density of a pixel inside a character. An object of the present invention is to provide an image processing device that detects characters with high accuracy by detecting characters based on the character. In order to achieve the above object, according to the present invention, a digital image signal is converted into predetermined first and second digital image signals.
Pixels with a density equal to or lower than the first threshold
A method for classifying pixels having a density equal to or higher than the second threshold to black pixels
And the connectivity between the white pixel and the black pixel based on the classification result.
Means for determining, and the determination of the pixel of interest based on the determination result of the connectivity.
Connected white pixels and connected black pixels in neighboring areas
When the elements are simultaneously included in a predetermined number or more,
Means for detecting the pixel of interest as a line drawing edge pixel.
First detecting means for detecting the digital image signal in a horizontal direction.
Means for differentiating the digital image signal in the vertical direction
Means for differentiating, the differential output in the horizontal direction and the vertical output
Compare the magnitude of the differential output, and determine the large differential output
When the pixel of interest is larger than the third threshold value of
A second detecting means comprising: means for detecting as
Line image edge pixels detected by the first detection means
Is determined to be a line drawing, or the line drawing error is detected by the first detecting means.
For a predetermined number of pixels following the detected pixel
The detection output of the second detection means is made valid,
Invalidating the detection output for the pixel,
When the detection output is valid, the second detection means
For determining a line drawing edge pixel detected as a line drawing
It is characterized by comprising further comprising a stage. An embodiment of the present invention will be specifically described below with reference to the drawings. FIG. 1 shows the configuration of an embodiment of the present invention. In the figure, 1 is an MTF correction circuit, 2 and 3 are binarization circuits, 4 and 5 are pattern matching circuits, 6, 7
Is a judgment circuit, 8 and 16 are AND circuits, 9 and 10 are differentiating circuits, 11 and 12 are absolute value circuits, 13 and 14 are comparators,
5 is an effective signal generator, 17 is an OR circuit, and 18 is an expansion circuit. The MTF correction circuit 1, the binarization circuits 2, 3, the pattern matching circuits 4, 5, the determination circuits 6, 7, and the AND circuit 8 constitute a first line drawing detection circuit. Characters are detected based on the absolute density of pixels. Further, a second line drawing detecting circuit is constituted by the differentiating circuits 9 and 10, the absolute value circuits 11 and 12, and the comparators 13 and 14, and detects a character based on the density difference between pixels. An MTF correction circuit 1 converts a digital signal obtained by reading a document with a scanner (not shown) into an MTF.
to correct. For example, a filtering process is performed using a 3 × 3 mask as shown in FIG. The signal after the MTF correction is input to the binarization circuits 2 and 3. The binarizing circuit 2 compares the signal after MTF correction with a threshold value th1, and determines that the pixel is a white pixel when the signal ≦ th1. The white pixel signal is input to the pattern matching circuit 4. The pattern matching circuit 4 performs pattern matching on the white pixel signal with a pattern (a pattern for examining the connectivity of white pixels) shown in FIG. The center pixel is determined as the active pixel. The determination circuit 6 includes a pattern matching circuit 4
The number of pixels determined to be an active pixel is counted in the local region including the target pixel, and if the count value is equal to or greater than a predetermined value, the target pixel or the local region including the target pixel is determined as a white line image pixel. Activate the output. The binarizing circuit 3 converts the input signal to a threshold value t.
Compared with h2, when the signal value ≧ th2, it is determined to be a black pixel. The black pixel signal is input to the pattern matching circuit 5. The pattern matching circuit 5 performs pattern matching on the black pixel signal with a pattern (a pattern for examining the connectivity of black pixels) shown in FIG. 4 using a 3 × 3 matrix. The center pixel is determined as the active pixel. The judgment circuit 7 includes a pattern matching circuit 5
The number of pixels determined to be an active pixel is counted in the local region including the target pixel, and if the count value is equal to or greater than a predetermined value, the target pixel or the local region including the target pixel is determined as a black line drawing pixel. Activate the output. The AND circuit 8 outputs a logical product signal of the outputs of the decision circuits 6 and 7. This output becomes active when the outputs of the determination circuits 6 and 7 are simultaneously active.
In other words, this is a case where both a white line image pixel and a black line image pixel exist in the local area including the target pixel, and the target pixel is detected as an edge pixel of the line image. The differentiating circuit 9 performs horizontal differentiation on the input image signal, and the differentiating circuit 10 performs vertical differentiation on the input image signal. FIG. 5 (a)
Indicates a horizontal differential filter, and (b) indicates a vertical differential filter. The reason why the differential processing is performed only in the horizontal direction and the vertical direction is that most of the kanji are likely to be crushed inside the character, and most of the kanji are composed of vertical lines and horizontal lines. The absolute value circuits 11 and 12 output the absolute values of the outputs of the differential filters 9 and 10, respectively. Comparator 13
Compares the outputs of the absolute value circuits 11 and 12, selects a large value as a result of the comparison, and outputs the selected value to the comparator 14. The comparator 14 compares the output from the comparator 13 with a threshold ths, and activates the output when the output of the comparator 13 is larger than ths. As shown in FIG. 6, for example, as shown in FIG. 6, the valid signal generator 15 holds the output signals of eight pixels from the OR circuit 17 in an 8-bit shift register, and outputs the logical sum of each bit from the OR. . That is, the MTF correction circuit 1
The output of the valid signal generation unit 15 becomes active in a section of eight pixels following the line drawing edge pixel detected by the first line drawing detecting circuit constituted by the ND circuit 8, and the second signal is generated within this valid section. When the output of the line drawing detection circuit (consisting of the comparator 14 to the differentiation circuit 9, 10) is active, AND
The output of circuit 16 becomes active. The outputs of the AND circuits 8 and 16 are connected to an OR circuit 17
By taking the logical sum, the detection output of the final line eh Tsu di pixels is obtained. Further, in the expansion circuit 18, a line drawing area is detected by expansion processing of line pixel edge pixels. For example, the expansion circuit 18 checks the presence or absence of a line image edge pixel for each 3 × 3 matrix, and if there is even one, determines the pixel of interest as a line image and activates the output. According to the present embodiment, a low-contrast crushed character or the inside of a small character can be detected as a character with higher precision without increasing the number of errors in determining a photograph or a halftone image as a line drawing. . As described above, according to the present invention, the following effects can be obtained. (1) Since the method of detecting a character based on the absolute density of a pixel and the method of detecting a character based on a density difference of a pixel are used together, a character can be accurately detected while compensating for the disadvantages of each method. It becomes possible to detect. (2) Since the means for judging an edge based on the absolute density of a pixel is always effective, erroneous judgment in a picture can be greatly reduced. (3) Since the means for judging an edge based on the density difference of pixels is effective only in the vicinity of a portion judged to be an edge by the means for judging an edge based on the absolute density of a pixel, It is possible to detect a low contrast portion inside a character as an edge without increasing erroneous determination. (4) Since not only the absolute density of pixels but also the continuity of black pixels and white pixels is determined, highly accurate character detection is possible. (5) Since the spatial differentiation is limited to the horizontal and vertical directions, the amount of hardware is suppressed, and efficient character detection suitable for image characteristics can be performed.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 shows a configuration of an embodiment of the present invention. FIG. 2 shows a filter for MTF correction. FIG. 3 shows a pattern for checking connectivity of white pixels. FIG. 4 shows a pattern for checking connectivity of black pixels. FIG. 5A shows a horizontal differential filter,
(B) shows a vertical differential filter. FIG. 6 shows a configuration of an effective signal generator. FIG. 7 is a diagram for explaining detection inside a character; [Description of Signs] 1 MTF correction circuit 2, 3 Binarization circuit 4, 5 Pattern matching circuit 6, 7 Judgment circuit 8, 16 AND circuit 9, 10 Differentiator circuit 11, 12 Absolute value circuit 13, 14 Comparator 15 Valid Signal generator 17 OR circuit 18 Expansion circuit

Claims (1)

(57) [Claims 1] A digital image signal is converted into predetermined first and second digital image signals.
A pixel having a density equal to or lower than the first threshold is compared with the threshold,
Means for classifying pixels having a density equal to or higher than a second threshold value to black pixels
And the connectivity between the white and black pixels is determined based on the classification result.
Means for determining the proximity of the pixel of interest based on the connectivity determination result.
Connected white pixels and connected black pixels in the side area
Note that when a specified number of
Means for detecting an eye pixel as a line drawing edge pixel
First detecting means; means for differentiating the digital image signal in the horizontal direction;
Means for differentiating the digital image signal in the vertical direction;
Ratio of the differential output in the horizontal direction and the differential output in the vertical direction
The differential output having a larger value is larger than a predetermined third threshold value.
Means for detecting the pixel of interest as a line drawing edge pixel
And a line image edge pixel detected by the first detecting means.
Is determined to be a line drawing, or the line is detected by the first detecting means.
A predetermined number of pixels following the pixels detected as image edge pixels
To enable the detection output of the second detection means,
Invalidating the detection output for an outside pixel,
When the detection output of the second means is valid,
The line image edge pixel detected by
An image processing apparatus comprising: