JPH05143774A - Method for correcting inclination of character document - Google Patents

Abstract

PURPOSE:To cancel the unexpected inclination by thinning the line while expanding the original picture for the optimum expansion time and rotating the picture to cancel the calculated inclination on the higher components from among the longitudinal and lateral components of each part. CONSTITUTION:The picture of the character document is successively expanded, and the area difference of the graphic before and after the expansion is calculated. When the ratio between the difference and the area becomes less than the prescribed value, the number of expansion one time less than the conventional expansion number is regarded as the optimum expansion number. Then the main component of the document, that is, whether it is the laterally or vertically written document is judged by unifying the inclination component of the pattern of each picture part. When (pattern number Nu in the vertical direction) <= (pattern number Nh in the lateral direction), the main component is judged to be the lateral direction. The picture is expanded for the optimum expansion number in the main component direction. The generated rectangle is thinned while expanding the picture for the optimum expansion number in the main component direction. The longitudinal and lateral component of each part of the thinning picture is calculated. The picture is rotated to cancel the inclination calculated based on the statistical amount in the higher components from among these components.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for correcting the inclination of a character original, and more particularly to eliminating an unexpected inclination of the original in a device such as a copying machine, a facsimile, an OCR, or the like which reads the set original as an image. The present invention relates to a tilt correction method.

[0002]

2. Description of the Related Art Due to the inclination of a character original, the data compression rate is lowered in a facsimile and the recognition rate is lowered in an OCR. Further, in the copying machine, in order to stabilize the copy quality, it is necessary to automatically correct the inclination of the document.

Conventionally, as inclination detection methods, expansion, contraction,
A method using thinning was known (Nakamura, Makoto Ujiie,
Noriyoshi Okamoto, Satoshi Minami, "Algorithm Extraction Algorithm for Character Area for Mixed Mode Communication", IEICE Transactions VO
L. J67-D, No. 11, pp1277-128
4, November 1984).

According to this method, a character original is binarized,
The characters are connected to each other by expansion. Further, the character string is displayed as a comparatively thin band by contraction and filling, and this is thinned. Also, the irregular portions at both ends of the thinned image are deleted by degeneration. In such a thinned image, after labeling, determine the inclination of a group having a length of a predetermined value or more,
This inclination is regarded as the original inclination. Here, the number of times of expansion is determined based on an empirical rule, and the optimum value could not be obtained unless the size of the text original, the quality, and the like were visually judged.

[0005]

SUMMARY OF THE INVENTION The present invention was devised to solve the above-mentioned conventional problems, and is capable of executing an expansion process most suitable for tilt detection, and has high detection accuracy.
An object of the present invention is to provide a method for correcting the inclination of a text original.

[0006]

According to the skew correction method for a text original according to the present invention, the image of the text original is successively expanded, the difference between the areas of the figures before and after expansion is calculated, and the ratio of the difference to the area is a predetermined value. When it becomes the following, the number of times of expansion that is one less than the number of times of expansion up to then is regarded as the optimum number of times of expansion. Then, the vertical and horizontal components of the contour of the figure are calculated for the initial image, only the direction of more components is expanded by the optimum expansion number, and the expanded image is thinned, and each part of the thinned image is expanded. Calculate the vertical and horizontal components, of these components,
For more components, the slope is calculated based on the statistical amount and correction is performed. According to the inclination correction method for a character original according to the present invention, each line of the character original can be processed as a substantially rectangular band, and by thinning it, the inclination can be detected and corrected with high accuracy.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a skew correction method for a text original according to the present invention will be described below with reference to the drawings. FIG. 1 is a flowchart showing the same embodiment. First, an original is read from an input means such as a scanner (step 1-1), and the size of the image is adjusted so as to be suitable for the subsequent processing (step 1).
-2). For example, the output image of the scanner is 4096 × 40
The number of pixels is 96 and the image to be processed by the image processing unit is 1024.
If it is × 1024, it is necessary to reduce it by ¼. In this reduction process, if the average value of the predetermined convolution is used as the density of the specific pixel, the grayscale information of the original image can be stored to the maximum extent. That is, in the case of 1/4 reduction, the average value of 4 × 4 convolution is set as the density of a specific pixel (for example, the upper left pixel in the same convolution).

Next, a density histogram of the image is obtained, binarization is performed by the optimum threshold value (step 1-3), and noise is further removed (step 1-4). As a method of calculating the optimum threshold value, a method of using the median density as a threshold value, a mode method
There are discriminant analysis method, P-tile method and the like. Also, as a noise removal method, isolated point removal, median filter and the like are effective. When detecting the inclination, the image of the character original is expanded in the main component direction, that is, the horizontal original is expanded in the horizontal direction, and the vertical original is expanded in the vertical direction an optimum number of times to convert each line into a band-shaped figure. By thinning this figure, the inclination can be easily detected. Here, the calculation of the optimum expansion number Ndi1 will be described (steps 1-5 to 1-13).

In step 1-5, the number Nwi of non-background pixels after noise removal is counted to obtain an initial value of the number of non-background pixels. Since the expansion is repeated thereafter, the loop counter i for counting the number of expansions is initialized (step 1-6). Then, expand once (step 1-7)
The loop counter i is incremented every time (step 1-8), the non-background pixel number Nw (i + 1) is counted, and the third derivative of the non-background pixel number is obtained (step 1-10). The third derivative is calculated by the following formula.

Third-order derivative Δ ³ Nwi = Δ ² Nw (i + 1) −Δ ² Nwi However, Δ ² Nwi = ΔNw (i + 1) -ΔNwi ΔNwi = Nw (i + 1) -Nwi

When the value of the ^third derivative Δ ³ Nwi becomes negative, interline fusion has already occurred, and the slowdown of the increase in the figure area due to expansion is detected (step 1-10).
Therefore, at this stage, the expansion loop is exited and the number of expansions up to the previous time, Ndil = i-1 is adopted as the optimum number of expansions (step 1-13). Then, the previous dilated image is used for the following main component detection (step 1-14). In this way, by tracing the change in the area of the figure for each expansion of one layer, the global fusion start can be surely detected, and the optimum number of expansions can be obtained.

It is possible that the start of fusion is unclear due to the influence of noise that cannot be completely removed by the noise removal. Therefore, the upper limit I of the number of expansions is set in advance, and even before the third derivative reaches a negative value,
The expansion is completed (step 1-11). In this case, the optimal number of expansions is (I-1). Next, the main component of the original, that is, whether the original is a horizontal original or a vertical original is determined. This judgment is made by summing up the inclination components of the pattern of each part of the image (step 1-14).

In calculating the inclination component, in the 3 × 3 convolution, the patterns of FIGS. 2A to 2L are horizontal components and the patterns of FIGS. Depending on whether there are many, the main component for the entire image is obtained. These patterns are the outline parts of the figure. The outline inclination is defined for each pixel row of two pixels, and the one in which the total value of the inclinations in the convolution is near horizontal is the horizontal pattern, and the one near vertical is the vertical pattern. Directional pattern. A pattern including one background pixel at the periphery of the convolution ((K) in FIGS. 2 and 3,
Regarding (L), a vertical pattern is used when the background pixels are on the left and right edges, and a horizontal pattern is used when the background pixels are on the upper and lower edges.

Assuming that the number of vertical patterns is Nv and the number of horizontal patterns is Nh, the main component is determined to be horizontal when Nv≤Nh, and is determined to be vertical if not (step 1 -15 to 1-17). Here, the index of the main component is horv, and in the vertical direction, h
Substitute "v" for orv and "h" for horv in the horizontal direction (steps 1-16, 1-17).

Here, the image is expanded a predetermined number of times in the direction of the main component to connect the characters in each line. The number of times of expansion should be the minimum number of times that characters are surely connected, and empirically, a value such as 9 Ndil / 4 [times] is adopted based on the Ndil. In this way, in the image generated by the optimum number of expansions, each line is a substantially rectangular figure. Next, the substantially rectangular figure is thinned. What is thinning?
By sequentially removing the pixels around the figure one by one, without changing the topology of the figure, the figure has a pixel width of 1
It is to be a line figure. As a thinning image processing method, for example, the method of HILDITCH, Japanese Patent Laid-Open No.
A method such as 4-13680 is effective. By this thinning, the almost center line of the substantially rectangular figure is obtained, and an effective element for calculating the inclination of the figure can be obtained. However, when thinning is performed, the end of the thinned graphic may be branched or curved as shown in FIG. 4C due to the shape of the characters at the end of the graphic or noise. Therefore, in order to remove these irregularities that hinder the accurate calculation of the inclination, degeneracy processing of the end points is performed after performing the thinning (FIG. 4D). As a result, the irregular portions at both ends are removed, and the accuracy of the subsequent slope calculation is increased.

It should be noted that the number of times of degeneracy should be the minimum number of times that an irregular portion can be completely erased. Empirically, a value such as 15 Ndil / 4 + 1 [times] is adopted based on the Ndil. It This empirical formula means erasing a line having a length corresponding to approximately 1 to 1.5 characters.

In calculating the inclination of the image mainly composed of the horizontal component, FIGS. 6 (A), (B), (C), (D), (E),
The numbers of the patterns (F), (G), (H), and (I) are counted, and the counted values are multiplied by a coefficient and summed. Here, the shape of each pattern and the numerical values to be multiplied by the number of each pattern are as follows.

Horizontal pattern shape and coefficient (A) 3 pixel horizontal pattern coefficient = 0 (B) (2 pixel right-down pattern) + (2 pixel horizontal pattern) coefficient = 0.5 (C) (2 pixel Horizontal pattern) + (2 pixel right-down pattern) Coefficient = 0.5 (D) (2 pixel horizontal pattern) + (2 pixel upper right pattern) Coefficient = -0.5 (E) (2 pixel right Upward pattern) + (horizontal pattern of 2 pixels) Coefficient = -0.5 (F) Right downward pattern of 3 pixels Coefficient = 1 (G) Right upper pattern of 3 pixels Coefficient = -1 (H) (right of 2 pixels) Downward pattern) + (2 pixel upper right pattern) coefficient = 0 (I) (2 pixel upper right pattern) + (2 pixel right down pattern) coefficient = 0

These coefficients are error components with respect to the horizontal component of each pattern, and give an absolute value "1" for a diagonal pattern of 3 pixels. These patterns (A) ~
The total number of images (I) is Na to Ni.
Then, the tilt angle θ is given by the following equation.

Θ = tan ^-1 [{0.5 (Nb + Nc) -0.5 (Nd + Ne) + Nf-Ng} / (Na + Nb +
Nc + Nd + Ne + Nf + Ng + Nh + Ni)] The same applies to the processing for a vertically written character original.
FIG. 5 shows the processing contents of the horizontal document corresponding to (A) to (D).
It shows in (A)-(D). Then, vertical patterns corresponding to FIGS. 6A to 6I are shown in FIGS. The shapes and coefficients of these patterns are as follows.

Vertical Pattern Shape and Coefficient (A) Vertical Pixel Pattern of 3 Pixels Coefficient = 0 (B) (Left-Down Pattern of 2 Pixels) + (Vertical Pattern of 2 Pixels) Coefficient = 0.5 (C) (of 2 Pixels) Vertical pattern) + (2-pixel left-down pattern) Coefficient = 0.5 (D) (2-pixel vertical pattern) + (2-pixel right-down pattern) Coefficient = -0.5 (E) (2-pixel right Downward pattern) + (2-pixel vertical pattern) Coefficient = -0.5 (F) 3 pixel right down pattern coefficient = -1 (G) 3 pixel upper right pattern coefficient = 1 (H) (2 pixel left Downward pattern) + (2-pixel right-down pattern) coefficient = 0 (I) (2-pixel right-down pattern) + (2-pixel left-down pattern) coefficient = 0

When the total number of the images of these patterns (A) to (I) is Na to Ni, the tilt angle θ is given by the following equation. θ = tan ⁻¹ [{0.5 (Nb + Nc) −0.5 (Nd
+ Ne) -Nf-Ng} / (Na + Nb + Nc + Nd +
Ne + Nf + Ng + Nh + Ni)]

In this way, an expanded figure obtained by performing the optimum number of expansions in the main component direction is generated, the expanded image is thinned, the vertical and horizontal components of this thinned image are calculated, and the larger one of these components is calculated. By detecting the slopes of the components in accordance with the statistic, it is possible to easily and accurately calculate the slopes. Next, in order to perform tilt correction, after the tilt angle is calculated, the image is affine-transformed so as to eliminate this tilt. Affine conversion can be executed by matrix calculation by software or address conversion processing by a dedicated IC.

[0024]

As described above, according to the inclination correction method for a text original according to the present invention, the image of the text original is successively expanded, and the optimum expansion number is obtained based on the change in the difference in the area of the graphic before and after expansion. Since the image that has been expanded is thinned,
It is possible to detect the inclination with high accuracy only by evaluating the thinned image, and since the image is rotated so as to eliminate the inclination, it is possible to surely execute the inclination correction and the processing is fast regardless of the number of groups. Has excellent effect.

[Brief description of drawings]

FIG. 1 is a flowchart showing an embodiment of a tilt detecting method according to the present invention.

FIG. 2A is a conceptual diagram showing a lateral component pattern including an error component of 0.5.

FIG. 2B is a conceptual diagram showing another lateral component pattern including an error component of 0.5.

FIG. 2C is a conceptual diagram showing another lateral component pattern including an error component of 0.5.

FIG. 2D is a conceptual diagram showing another lateral component pattern including an error component of 0.5.

FIG. 2 (E) is a conceptual diagram showing another lateral component pattern including an error component of 0.5.

FIG. 2F is a conceptual diagram showing another lateral component pattern including an error component of 0.5.

FIG. 2G is a conceptual diagram showing another lateral component pattern including an error component of 0.5.

FIG. 2H is a conceptual diagram showing another lateral component pattern including an error component of 0.5.

FIG. 2 (I) is a conceptual diagram showing a horizontal lateral component pattern.

FIG. 2 (J) is a conceptual diagram showing another horizontal lateral component pattern.

FIG. 2 (K) is a conceptual diagram showing a lateral component pattern including one background pixel.

FIG. 2 (L) is a conceptual diagram showing another lateral component pattern including one background pixel.

FIG. 3A is a conceptual diagram showing a vertical component pattern including an error component of 0.5.

FIG. 3B is a conceptual diagram showing another vertical component pattern including an error component of 0.5.

FIG. 3C is a conceptual diagram showing another vertical component pattern including an error component of 0.5.

FIG. 3D is a conceptual diagram showing another vertical component pattern including an error component of 0.5.

FIG. 3E is a conceptual diagram showing another vertical component pattern including an error component of 0.5.

FIG. 3F is a conceptual diagram showing another vertical component pattern including an error component of 0.5.

FIG. 3G is a conceptual diagram showing another vertical component pattern including an error component of 0.5.

FIG. 3 (H) is a conceptual diagram showing another vertical component pattern including an error component of 0.5.

FIG. 3 (I) is a conceptual diagram showing a vertical vertical component pattern.

FIG. 3 (J) is a conceptual diagram showing another vertical vertical component pattern.

FIG. 3 (K) is a conceptual diagram showing a vertical component pattern including one background pixel.

FIG. 3 (L) is a conceptual diagram showing another vertical component pattern including one background pixel.

FIG. 4A is a conceptual diagram showing an example of a horizontal document.

FIG. 4B is a conceptual diagram showing a laterally expanded figure of a horizontally written document.

FIG. 4 (C) is a conceptual diagram showing a thinned image of a horizontally expanded document of a horizontally written document.

FIG. 4D is a conceptual diagram showing a state in which a thinned image of a horizontally expanded document of a horizontally written document is subjected to a shrinking process.

FIG. 5A is a conceptual diagram showing an example of a vertically written document.

FIG. 5B is a conceptual diagram showing an expanded graphic in the vertical direction of a vertically written document.

FIG. 5C is a conceptual diagram showing a thinned image of an expanded figure in the horizontal direction of a vertically written document.

FIG. 5D is a conceptual diagram showing a state in which a thinned image of an expanded figure in a horizontal direction of a vertically written document has been subjected to degeneration processing.

FIG. 6A is a conceptual diagram showing a horizontal pattern.

FIG. 6B is a conceptual diagram showing a pattern including an error component of 0.5 with respect to a horizontal pattern.

FIG. 6C is a conceptual diagram showing another pattern including an error component of 0.5 with respect to the horizontal pattern.

FIG. 6D is a conceptual diagram showing another pattern including an error component of 0.5 with respect to the horizontal pattern.

FIG. 6E is a conceptual diagram showing still another pattern including an error component of 0.5 with respect to the horizontal pattern.

FIG. 6 (F) is a conceptual diagram showing a right-down pattern.

FIG. 6G is a conceptual diagram showing an upper right pattern.

FIG. 6 (H) is a conceptual diagram showing a pattern that changes from downward rightward to upward rightward.

FIG. 6 (I) is a conceptual diagram showing a pattern changing from upper right to lower right.

FIG. 7A is a conceptual diagram showing a vertical pattern.

FIG. 7B is a conceptual diagram showing a pattern including an error component of 0.5 with respect to a vertical pattern.

FIG. 7C is a conceptual diagram showing another pattern including an error component of 0.5 with respect to the vertical pattern.

FIG. 7D is a conceptual diagram showing still another pattern including an error component of 0.5 with respect to the vertical pattern.

FIG. 7E is a conceptual diagram showing still another pattern including an error component of 0.5 with respect to the vertical pattern.

FIG. 7 (F) is a conceptual diagram showing a pattern on the right.

FIG. 7G is a conceptual diagram showing an upper right pattern.

FIG. 7 (H) is a conceptual diagram showing a pattern changing from leftward descending to rightward descending.

FIG. 7 (I) is a conceptual diagram showing a pattern changing from rightward down to leftward.

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] October 27, 1992

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be corrected] Brief description of the drawing

[Correction method] Change

[Correction content]

[Brief description of drawings]

FIG. 1 is a flowchart of steps 1-1 to 1-13 showing an embodiment of a method for correcting inclination of a text original according to the present invention.

FIG. 2 is a conceptual diagram for explaining a horizontal component pattern. (A) It is a conceptual diagram which shows the horizontal component pattern containing an error component of 0.5. (B) is a conceptual diagram showing another lateral component pattern including an error component of 0.5. (C) is a conceptual diagram showing another lateral component pattern including an error component of 0.5. (D) is a conceptual diagram showing another lateral component pattern including an error component of 0.5. (E) is a conceptual diagram showing another lateral component pattern including an error component of 0.5. (F) is a conceptual diagram showing another lateral component pattern including an error component of 0.5. (G) is a conceptual diagram showing another lateral component pattern including an error component of 0.5. (H) is a conceptual diagram showing another lateral component pattern including an error component of 0.5. (I) It is a conceptual diagram which shows a horizontal lateral component pattern. (J) It is a conceptual diagram which shows another horizontal lateral component pattern. (K) It is a conceptual diagram which shows the horizontal component pattern containing one background pixel. (L) is a conceptual diagram showing another lateral component pattern including one background pixel.

FIG. 3 is a conceptual diagram for explaining a vertical component pattern. (A) It is a conceptual diagram showing a vertical component pattern including an error component of 0.5. (B) is a conceptual diagram showing another vertical component pattern including an error component of 0.5. (C) is a conceptual diagram showing another vertical component pattern including an error component of 0.5. (D) is a conceptual diagram showing another vertical component pattern including an error component of 0.5. (E) is a conceptual diagram showing another vertical component pattern including an error component of 0.5. (F) It is a conceptual diagram showing another vertical component pattern including an error component of 0.5. (G) is a conceptual diagram showing another vertical component pattern including an error component of 0.5. (H) is a conceptual diagram showing another vertical component pattern including an error component of 0.5. (I) It is a conceptual diagram which shows a vertical vertical component pattern. (J) It is a conceptual diagram which shows another vertical vertical component pattern. (K) It is a conceptual diagram which shows the vertical component pattern containing one background pixel. (L) is a conceptual diagram showing another vertical component pattern including one background pixel.

FIG. 4 is a conceptual diagram for explaining a horizontal document. FIG. 3A is a conceptual diagram showing an example of a horizontally written document. FIG. 9B is a conceptual diagram showing a laterally expanded figure of a horizontally written document. (C) is a conceptual diagram showing a thinned image of a horizontally expanded document of a horizontally written document. FIG. 6D is a conceptual diagram showing a state in which a thinned image of a horizontally expanded document of a horizontally written document is subjected to degeneration processing.

FIG. 5 is a conceptual diagram for explaining a vertically written document. FIG. 3A is a conceptual diagram showing an example of a vertically written document. FIG. 6B is a conceptual diagram showing a vertically expanded figure of a vertically written document. (C) is a conceptual diagram showing a thinned image of a horizontally expanded figure of a vertically written document. FIG. 6D is a conceptual diagram showing a state in which a thinned image of a horizontally expanded figure of a vertically written document is subjected to degeneration processing.

FIG. 6 is a conceptual diagram for explaining a horizontal pattern. (A) It is a conceptual diagram which shows a horizontal pattern. (B) is a conceptual diagram showing a pattern including an error component of 0.5 with respect to a horizontal pattern. (C) is a conceptual diagram showing another pattern including an error component of 0.5 with respect to the horizontal pattern. (D) is a conceptual diagram showing another pattern including an error component of 0.5 with respect to the horizontal pattern. (E) is a conceptual diagram showing still another pattern including an error component of 0.5 with respect to the horizontal pattern. (F) It is a conceptual diagram which shows the pattern of right descent. (G) It is a conceptual diagram which shows the pattern of upper right. (H) It is a conceptual diagram which shows the pattern which changes from a right descent to an upper right. (I) It is a conceptual diagram which shows the pattern which changes from the upper right to the lower right.

FIG. 7 is a conceptual diagram for explaining a vertical pattern. (A) It is a conceptual diagram which shows a vertical pattern. (B) is a conceptual diagram showing a pattern including an error component of 0.5 with respect to a vertical pattern. (C) is a conceptual diagram showing another pattern including an error component of 0.5 with respect to the vertical pattern. (D) is a conceptual diagram showing still another pattern including an error component of 0.5 with respect to the vertical pattern. (E) is a conceptual diagram showing still another pattern including an error component of 0.5 with respect to the vertical pattern. (F) It is a conceptual diagram which shows the pattern of right descent. (G) It is a conceptual diagram which shows the pattern of upper right. (H) It is a conceptual diagram which shows the pattern which changes to the right descending from the left descending. (I) It is a conceptual diagram which shows the pattern which changes from a right descent to a left descent.

FIG. 8 is a flowchart of steps 1-14 to 1-23.

[Procedure Amendment 2]

[Document name to be corrected] Drawing

[Correction target item name] All drawings

[Correction method] Change

[Correction content]

[Figure 1]

[Figure 4]

[Figure 5]

[Figure 6]

[Figure 7]

[Figure 8]

[Fig. 2]

[Figure 3]

Claims (1)

[Claims] 1. An image of a text original is successively expanded, a difference in area between figures before and after expansion is calculated, and when the ratio of the difference to the area is less than a predetermined value, the number of expansions is greater than the number of expansions up to that point. 1
The number of times of less expansion is regarded as the optimal number of expansions, the vertical and horizontal components of the contour of the figure are calculated for the initial image, the expansions of the optimal number of expansions are performed only in the direction of more components, and the thinned image is subjected to the thinning processing. The vertical and horizontal components of each part of this thinned image are calculated, the inclination of more components is calculated based on statistics, and the image is rotated so as to eliminate this inclination. A method for correcting the skew of a text original characterized by.