JPH0390980A - Method and device for detecting tilt of document picture - Google Patents

Abstract

PURPOSE:To detect the tilt of a document picture in a stable and easy way by detecting the direction where the projection, i.e, the width of the black picture element showing a character row and a table frame when the character row and a table, etc., are projected in the tilt direction. CONSTITUTION:A pattern memory 1 stores the document pictures as the quantized picture information. A binary projection information extracting part 2 sets (m) pieces of partial information (i) of a fixed size in the prescribed direction, scans each partial area (i), and extracts the binary projection information showing the presence/absence of black picture elements on each scan line to store them in sequence into a binary projection information storage part 3. Then the adjacent binary projection information are shifted in steps in the vertical and horizontal shift direction by a prescribed extent and read out. At the same time, an OR is obtained among plural binary projections and the cumulative binary projection information are extracted. Then the shift value is changed for detection of the shift value and the shift direction where the number of black picture elements of the cumulative binary projection information is minimized. Thus the tilt of a document picture can be detected in an easy and stable way.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method and apparatus for detecting the skew of a document image manually generated by an image input device.

[Prior art] In order to automatically read general printed documents, etc., first, a general printed document, etc.
It is necessary to capture the document image as a quantized document image, but since it is tilted during input, it is normal that it is tilted somewhat with respect to the horizontal or vertical scanning direction of a human image. For this reason, in documents with narrow spacing between character lines, the character lines may not be detected, or the resolution of the human-powered image device must be increased, resulting in performance deterioration. There is a need to.

Conventionally, in such a document image, skew detection is performed using the fact that characters are regularly arranged in a text area at the center of the document image.

For example, "Paper component extraction method that does not depend on format specification information" (IEICE, Journal of Electronics, Information and Communication Engineers, v.

L, J66-D, No. 1. PPIII~118°19
As shown in 83), a first method is known in which a histogram is created in the direction of angle θ while changing the inclination angle θ sequentially, and the angle θ at which peaks and valleys on the histogram appear conspicuously is determined as the inclination angle. ing.

In addition, as shown in the above document, the input document image is divided into multiple band-shaped regions, a histogram is created for each pixel, and the tilt angle θ is calculated from the correlation value for the histogram distribution between each adjacent region. A second method of determining this is known.

[Problem to be solved by the invention]

However, in the first method, it is necessary to calculate a histogram every time the angle θ direction changes, which poses a problem in terms of processing time.

In addition, in the second method, the tilt angle θ is determined by corresponding to the peaks and valleys of the histogram (corresponding to text lines and line spacing) for each adjacent strip, so if a line of characters whose size changes is included, the tilt Detection accuracy deteriorates.

Furthermore, when using only the regular arrangement of character lines as in the first and second methods above, if areas such as tables are included, the tilt detection accuracy deteriorates, resulting in cost performance problems. occurs.

SUMMARY OF THE INVENTION Therefore, in order to solve the above-mentioned conventional problems, the layout of the entire document is arranged regularly on a rectangle. In other words, by utilizing the fact that the width of black pixels representing character lines and table frames is the minimum, it is possible to stably and easily detect the skew of a document image even if it includes character lines or tables with varying character sizes. An object of the present invention is to provide a document image tilt detection device that enables the following.

Another object of the present invention is to divide an input document image into a plurality of belt-shaped regions, extract binary projection information that detects only the presence or absence of black pixels in a predetermined scanning direction for each region, and extract the binary projection information for each region. An object of the present invention is to provide an apparatus for detecting the inclination of a sentence M image in which the processing speed is improved by detecting the inclination based on the cumulative binary projection information obtained by taking the logical sum of the information while shifting the information in a stepwise manner.

(Means for Solving the Problems) In order to solve the above-mentioned problems, the document image skew detection method according to the present invention is a method for detecting the skew of a binary quantized document image. The document image is scanned horizontally or vertically for each partial area of a plurality of band-shaped partial areas, and it is checked whether or not there are black pixels on each scanning line. extract the binary projection information indicating the presence or absence of the binary projection information, read out the adjacent binary projection information by shifting it by a predetermined amount in the vertical or horizontal shifting direction in a stepwise manner, and calculate the logical sum of the plurality of binary projections. the number of black pixels of the cumulative binary projection information is determined by changing the amount of shift in the vertical or horizontal direction, and the amount of shift and direction of shift that minimizes the number of black pixels; Detect.

In addition, the document image tilt detection device of the present invention includes a memory for storing a binary quantized document image, and a plurality of partial areas set in the form of bands of a constant size in the vertical or horizontal direction, so that the document image is horizontally quantized. Alternatively, a binary projection extracting means scans in the vertical direction and detects and detects the presence or absence of black pixels on each scanning line, and the adjacent two projection information is systed according to the vertical or horizontal shift direction and predetermined shift amount. cyst calculation means, cumulative binary projection information extraction means for calculating the logical sum of the plurality of binary projection information shifted in a stepwise manner, and black pixel number calculation means for detecting the number of black pixels of the cumulative binary projection information. means, a counter for changing the shift amount and shift direction, and means for detecting a minimum value of the number of black pixels of the plurality of cumulative binary projection information obtained by changing the shift amount and shift direction. Be equipped.

Further, the document image tilt detection device of the present invention includes a memory for storing a binary quantized document image, and a plurality of partial regions having a constant size in the vertical or horizontal direction. A binary projection extraction means that scans in the horizontal or vertical direction and detects the presence or absence of black pixels on each scanning line, and a cyst that performs cysts on the adjacent binary projection information according to vertical or horizontal shifting directions and predetermined shifting amounts. a calculating means, a cumulative binary projection information extracting means for calculating the logical sum of the plurality of binary projection information shifted in a stepwise manner, and a black pixel number calculating means for detecting the number of black pixels of the cumulative binary projection information. , a counter for changing the shift amount, and detecting the minimum value of the number of black pixels of the cumulative binary projection information calculated for two shift directions such as up and down when the shift amount is 1, and determining the shift direction. means for determining the number of black pixels of the cumulative binary projection information obtained by sequentially increasing the l'llJ displacement amount starting from 2 with respect to the determined shifting direction; and means for checking whether the number of black pixels S is greater than the number of black pixels in the binary projection information and interrupting the increase in the amount of shift.

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings.

Figure 1 shows a partial area obtained by dividing a tilted document image into two vertical bands of a certain size.
FIG. 3 is a diagram explaining the principle of obtaining binary projection information indicating the presence or absence of black pixels by scanning in the horizontal direction.

In FIG. 1(a), characters and lines within the document image are indicated by diagonal lines. 0 in the figure. 2. 4...24 is
Each horizontal scan line is shown, with arrows indicating the distance to the position of the first detected black pixel on each scan line.

FIG. 1 et al.) show binary projection information obtained for FIG. 1(a). That is, scanning line 4 in FIG. 1(a)
In this case, there is a black pixel indicating the character part, so the image shown in Fig. 1(b)
"1" is stored in scanning line 4 of the scanning line IO.
Since there is no black pixel in , the scan line 10 of FIG. 1 is stored. In this way, for each scan line, a binary number consisting of either "1" or "0" indicating the presence of a black pixel.
Value projection information is obtained.

FIG. 2 is a diagram shown to explain the principle of the tilt detection method according to the present invention.

FIG. 2(a) shows a tilted document image, in which three partial areas i with a constant size Δm are arranged in the vertical direction as indicated by dotted lines in the figure.
It shows how it is divided into (i=0.l, 2).

Figure 2b) shows the three partial areas shown in Figure 2(a) =0. 1. For 2), the binary projection information shown in FIG. 1 is shown, and the areas filled in with black in the figure indicate the presence of black pixels.

Tilt detection is performed as follows. Shift amount Δt in the vertical direction with respect to partial area 0 (Δt=0°1, 2.
3...), and set the reading position of the binary projection information of partial area 1 by △t in the vertical direction, and the reading position of the binary projection information of partial area 2 by 2Δt in the vertical direction (i.e. ,
By shifting the partial area 1 by an amount i/Δt) and calculating the logical sum of the three binary projection information, the cumulative binary projection information obtained by superimposing the binary projection information of the three partial areas is obtained. To detect.

FIG. 2(C) shows cumulative binary projection information obtained for the binary projection information of each partial area shown in FIG. 2(5).

FIG. 2C shows a case where the shift amount Δt=Q and a case where the read position of the binary projection information of the partial area 1 is shifted downward by Δt.

Tilt detection is performed by determining the above-mentioned shift amount and shift direction (upward or downward).

That is, what is necessary is to find the direction and amount of shift that minimizes the number of black pixels in the cumulative binary projection information obtained when shifting upward or downward by Δt. For example, in Figure 2 (C),
The number of black pixels in the cumulative binary projection information is smaller when shifted downward by Δt (corresponds to the area filled with black in the figure)
I understand that.

Here, in the first invention, a certain lower tolerance value is set in advance, and the shift amount Δt in both the upward and downward directions is set to 0.1.
2...T, the cumulative binary projection information is extracted for each, and the minimum shift amount and shift direction are determined from among the black pixel numbers.

In addition, in the second invention, in order to improve the processing speed of the first invention, firstly, the shift amount Δ is determined in the upward and downward directions.
At t=1, the number of black pixels in the cumulative binary projection information is compared to determine the shifting direction.

Next, when the direction of shift is determined, the amount of shift Δt=0°1.2
．． . . . are sequentially increased, and when the number of black pixels of the previously extracted cumulative binary projection information becomes smaller, the shift amount Δt is determined.

FIG. 3 is a functional block diagram showing an embodiment of the first invention. The button memory 1 stores document images as quantized image information.

As explained in FIGS. 1 and 2(a), the binary projection information extraction unit 2 extracts m partial regions i (i=o, 1.2...) of a constant size in a predetermined direction. m-1>, each partial region l is scanned, binary projection information indicating the presence or absence of a black pixel on each scanning line is extracted, and sequentially stored in the binary projection information storage unit 3.

The partial area counter 5 is the first partial area (i=o) stored in the binary projection information storage unit 3.

1.2. . . . m −1), and when the binary projection information of the partial area l is sequentially processed by the cyst calculation unit 4 and transferred to the logical sum calculation unit 9, it is counted up. Note that the maximum value of the partial area counter 5 is assumed to be m-'.

The shift amount counter 6 represents the shift amount Δt as explained in FIG. 2 et al. The shift direction register 8 indicates the shift direction (for example, upward or downward), as explained in FIG. 2A).

Partial area 1 (i=0
．． 1. 2. When the binary projection information of .
t> is calculated by the multiplier 7, and the binary projection information of the partial area l is shuffled according to the output value of the multiplier 7 and the contents of the shift direction register 8, and is transferred to the logical OR calculation unit 9.

The logical sum calculation unit 9 stores cumulative binary projection information 1. ! Part 10
The logical sum of the cumulative binary projection information stored in the storage unit 10 (however, the initial value is θ) and the output value to the cyst calculation unit 4 is calculated and transferred to the cumulative binary projection information storage unit 10.

Similarly, m partial regions i (i=0.1゜2
, . . . m-1) are sequentially transferred to the cyst calculation unit 4, and are stored in the cumulative binary projection information storage unit 10 via the logical sum calculation unit 9. The shift amount △ which is the content of the counter 6 and the shift direction register 8
Cumulative binary projection information as shown in FIG. 2(C) with respect to the shifting direction, which is the content of , is stored.

Next, the contents of the cumulative binary projection information storage section 10 are read out, and the number of black pixels of the cumulative binary projection information described above is calculated by the black pixel number calculation section 11 and transferred to the comparison section 12 .

The comparison unit 12 compares the minimum black pixel number (the initial value is a very large value) stored in the minimum black pixel number storage unit 13 and determines that the output value of the black pixel calculation unit 11 is the minimum black pixel number. If it is smaller than the content of the number storage unit ↓3, the output value of the black pixel calculation unit 11 is transferred to the minimum black pixel number storage unit 13, and the content of the shift amount counter 6 and the shift value are transferred by the control unit (not shown). The contents of the direction register 8 are transferred to the shift amount storage section 14 and the shift direction storage section 14, respectively.

Next, the content of the shift amount counter 6 is incremented by one, and the above-described operations are repeated until the content of the shift amount counter 6 reaches a certain allowable value T.

When the content of the shift amount counter 6 reaches a certain allowable value, the shift direction is changed by changing the content of the shift direction register 8 (for example, from 0 to 1), and the shift direction is changed (for example, downward (upwards) Repeat the above-mentioned operations.

As a result, the shift direction storage section 14 and the shift amount storage section 1
5, the direction and amount of shift are determined, and the tilt can be easily detected based on these.

FIG. 4 is a functional block diagram showing an embodiment of the second invention.

The button memory l stores a document image as quantized image information.

As explained in FIGS. 1 and 2(a), the binary projection information extraction unit 2 extracts m partial regions i (i=0. 1. 2.--) of a constant size in a predetermined direction. m-'), each partial area 1 is scanned, binary projection information indicating the presence or absence of a black pixel on each scanning line is extracted, and sequentially stored in the binary projection information storage section 3.

The partial area counter 5 calculates the first partial area (i=0.1°2,...) stored in the binary projection information storage section 3.
.
The count is incremented when the data is transferred to . Note that the maximum value of the partial area counter 5 is assumed to be m-'.

The shift amount counter 6 represents the shift amount Δt as explained in FIG. 2(b). The shift direction register 8
As explained in Figure (b), it represents the shifting direction (for example, upward or downward).

Partial area 1 (i=0
．． 1. 2.・−・−m−′) (7) When the binary projection information is transferred to the cast calculation unit 4, the product of the contents of the partial area counter 5 and the contents of the shift amount counter 6 (i.e., 1・
Δt) is calculated by the multiplier 7, so that the partial area 1
The binary projection information is sourced according to the output value of the multiplier 7 and the contents of the shift direction register 8, and is transferred to the OR operation B9.

The logical sum calculation unit 9 calculates the cumulative binary projection information stored in the cumulative binary projection information storage unit 10 (however, the initial value is 0).

) and the output value to the cyst calculation unit 4 is calculated and transferred to the cumulative binary projection information storage unit 10.

Similarly, m partial regions i (i=0.1゜2
, . . . m-1) are sequentially transferred to the cyst calculation unit 4, and are stored in the cumulative binary projection information storage unit 1O through the logical sum operation 89. 6 and shift direction register 8.
Cumulative binary projection information as shown in FIG. 2(C) with respect to the shifting direction, which is the content of , is stored.

The above functions are equivalent to those of the first invention shown in FIG.

Next, the contents of the cumulative binary projection information storage section IO are read out, the black pixel number calculation section 11 calculates the number of black pixels of the cumulative binary projection information, and the result is transferred to selection 921 .

The selection section 21 selects whether the output value of the black pixel number calculation section 11 is to be transferred to the minimum value calculation section 22 or the comparison section 24 depending on the contents of the shift amount counter 6.

That is, as explained in the case where the content of the shift amount counter 6 is 1 or less (for example, 0:l in FIG. 2), the shift amount Δt=
The number of black pixels in the cumulative binary projection information is transferred to the minimum value calculation unit 22 only for three types: Q, upward and downward shift amount Δt=1).

Here, the above three cases are executed in advance by the control unit (not shown).

The minimum value calculation unit 22 detects the minimum value of the three numbers of black pixels S and transfers it to the minimum black pixel storage unit 23, and the control unit stores the corresponding shift amount Δt in the shift amount storage unit 15. Further, the shifting direction is transferred to the shifting direction storage section 14 and the shifting direction register 8. Note that when the minimum value calculation unit 22 detects the case where the shift amount Δt=Q, the shift amount Δt is transferred to the shift amount storage unit 15, and the operation ends.

Next, the content of the shift amount counter, the shift amount △t=2
Similarly, when the number of black pixels of the cumulative binary projection information is detected, the shift amount Δt is set through the selection unit 21.
The number of black pixels for =2 is transferred to the comparison section 24 and compared with the contents of the minimum black pixel storage section 23.

In the comparison section 24, if the content of the minimum black pixel storage section 23 is larger than the value inputted to the comparison section 24 through the selection section 21, the value is transferred to the minimum black pixel storage section 23, and the shift amount counter 6 is The contents are transferred to the shift amount storage section 15, the value of the shift amount counter 6 is incremented by 1, and the same operation is repeated.

On the other hand, in the comparison section 24, if the content of the minimum black pixel storage section 23 is smaller than the value inputted to the comparison section 24 through the selection section 21, the operation ends.

Note that even if the value of the shift amount counter 6 exceeds a certain allowable value T, the above operation is terminated.

When the above operation is completed, the shift direction and shift amount are stored in the shift direction storage section 14 and the shift amount storage section 15, respectively.

〔Effect of the invention〕

As described above, according to the document image skew detection method and apparatus of the present invention, even if a document image contains text lines or tables whose font sizes change, even if the document image is skewed, it can be easily detected. Moreover, since tilt detection can be performed stably, cost performance can be significantly improved.

[Brief explanation of drawings]

Fig. 1 is a diagram for explaining two-residue projection information obtained by horizontally scanning a partial area on a document image input at an angle to check the presence or absence of black pixels. , FIG. 3 is a functional block diagram showing an embodiment of the first invention, and FIG. 4 is a functional block diagram showing embodiments of the second invention. It is a block diagram. 1...Bang memory, 2...Binary projection information extraction section, 3...Binary projection information storage section, 4.
...Instal calculation section, 5...Partial area counter, 6...Shift amount counter, 7...
...Multiplication section, 8...Shift direction register, 9.
. . . Logical sum calculation unit, 10 . . . Cumulative binary projection information storage unit, 11 . . . Black pixel number calculation unit, 12
... Comparison section, 13 ... Minimum black pixel number storage section, 14 ... Shift direction storage section, 15 ...
...Shift amount storage section, 21... Selection section, 22
... Minimum value calculation section, 23 ... Minimum black pixel storage section, 24 ... Comparison section.

Claims (3)

[Claims] (1) In a method for detecting the inclination of a binary quantized document image, the document image is scanned in the horizontal or vertical direction for each partial area of a plurality of partial areas in the form of bands of a constant size in the vertical or horizontal direction. Then, binary projection information indicating the presence or absence of a black pixel on each scanning line is extracted, and the adjacent binary projection information is read out by shifting it by a predetermined amount in the vertical or horizontal shifting direction, and The cumulative binary projection information is extracted by calculating the logical sum of the binary projections, and the number of black pixels of the cumulative binary projection information is determined by changing the vertical or horizontal shift amount. 1. A method for detecting a tilt of a document image, the method comprising: detecting a shift amount and shift direction that minimizes the shift amount and shift direction. (2) A memory for storing a binary quantized document image and a plurality of partial regions of a constant size in the vertical or horizontal direction are set, and the document image is scanned in the horizontal or vertical direction, and each binary projection extraction means for detecting the presence or absence of black pixels on a scanning line; cyst calculation means for cysting the adjacent binary projection information according to the vertical or horizontal shifting direction and a predetermined shifting amount; Cumulative binary projection information extraction means for calculating a logical sum of a plurality of binary projection information;
a black pixel number calculation means for detecting the number of black pixels of the cumulative binary projection information; a counter for changing the shift amount and the shift direction; and a plurality of the cumulative numbers obtained by changing the shift amount and shift direction. A document image tilt detection device comprising: means for detecting a minimum value of the number of black pixels in binary projection information. (3) A memory for storing a binary quantized document image and a plurality of partial regions of a fixed size in the vertical or horizontal direction are set, and the document image is scanned in the horizontal or vertical direction, and each binary projection extraction means for detecting the presence or absence of black pixels on a scanning line; cyst calculation means for cysting the adjacent binary projection information according to a vertical or horizontal shifting direction and a predetermined shifting amount; cumulative binary projection information extraction means for calculating a logical sum of the plurality of binary projection information; logical prime number calculating means for detecting the number of black pixels of the cumulative binary projection information; and a counter for changing the shift amount. and,
means for detecting a minimum number of black pixels of the cumulative binary projection information calculated for two shift directions such as up and down when the shift amount is 1, and determining a shift direction; and the determined shift. Checking whether the number of black pixels of the cumulative binary projection information obtained by sequentially increasing the amount of shift with respect to the direction is larger than the number of black pixels of the cumulative binary projection information that has already been detected; A document image tilt detection device, comprising: means for interrupting the increase in the amount of shift.