JPH0589284A - Device for discriminating longitudinal and transverse direction of document - Google Patents

Abstract

PURPOSE:To discriminate whether a character picture is written vertically or horizontally by subjecting projection distributions in X-axis and Y-axis directions of the document picture to one-dimensional Fourier transform and comparing values of obtained power spectrums with each other. CONSTITUTION:An inputted picture signal is converted to a binarized picture. Projection distribution in vertical and horizontal directions are obtained with respect to a divided area A of the document picture. Obtained projection distributions hX(x) and hY(y) are subjected to the one-dimensional Fourier transform of a discrete value system to obtain respective harmonic components (power spectrums). Values of these power spectrums are compared with each other to discriminate whether the input picture is written vertically or horizontally by the comparison result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention is used in OCR and the like.
The present invention relates to a document vertical / horizontal direction determination device for determining the vertical / horizontal direction of an input document image.

[0002]

2. Description of the Related Art When recognizing characters in a Japanese document image, it is necessary to first determine the vertical and horizontal directions.
As a layout analysis method therefor, there is a case in which vertical writing or horizontal writing is determined from a peak value obtained by binarizing a document image to obtain a projection distribution and further performing two-dimensional Fourier transform.

[0003]

By the way, since the above-mentioned conventional determination method uses the two-dimensional Fourier transform, the calculation processing amount becomes enormous and the processing time until the determination becomes long, which is not practical. There is. The present invention has been made to solve the above problems, and an object of the present invention is to provide a practical document vertical / horizontal direction determination apparatus capable of performing determination in a short time.

[0004]

In order to achieve the above object, the present invention provides a means for converting an input image signal into a binary image and a projection of the binary image in the X-axis and Y-axis directions. A means for obtaining a distribution, a means for obtaining a power spectrum by performing a one-dimensional Fourier transform of a discrete value system on each projection distribution,
It is characterized by comprising means for comparing the values of the respective power spectra and judging from which of the magnitudes the input image is in the vertical writing direction and the horizontal writing direction.

[0005]

In the present invention, the image signal of the input document image is converted into a binarized image, and the projection distributions of the binarized image in the X-axis and Y-axis directions are obtained. Then, for each projection distribution, 1 in the discrete value system
The dimensional Fourier transform is performed to obtain the power spectrum, and the magnitudes of the values are compared to determine whether the input image is in vertical writing or horizontal writing.

[0006]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is an explanatory view of a document image input to the embodiment, and FIG. 2 shows one of the divided image areas in FIG. 1 and shows it along with its projection distribution. In FIG. 1, a document image 1 is a binarized image obtained by binarizing an input image signal with a preset threshold value, and the circles in the document image 1 indicate individual images. It is the character 2. The characters 2 are continuous in the horizontal direction to form a line for horizontal writing. Further, in consideration of the case where the document has an inclination, the document image 1 is divided in the X and Y directions and processed for each of the areas A to D. The length of one side of the area is the number of pixels Wx and Wy, respectively.
Is.

In FIG. 2, projection distributions in the vertical and horizontal directions are obtained for the divided area A of the document image 1, and the respective distributions are h _x (x) and h _y (y). The projection distributions h _x (x) and h _y (y) obtained here are subjected to a one-dimensional Fourier transform in a discrete value system, and each harmonic component (power spectrum) C _{Find xn} and C _ym . Note that n and m are orders.

[0008]

[Equation 1]

[0009]

[Equation 2]

Next, the maximum values C _xmax and C _ymax are obtained from the harmonic components C _xn and C _ym , respectively, and the maximum values C _max = max [C _xmax and C _ymax ] of both are obtained. Area A
In the above, clear peaks and valleys do not appear in the projection distribution h _x (x) about the X axis, but peaks and valleys appear in the projection distribution h _y (y) about the Y axis. Therefore, C _ymax is selected as the maximum value C _max . With this, the processing for the area A is completed, and subsequently, the other areas B, C, and D are similarly processed.

The maximum value of each area A to D is C
_Amax , C _Bmax , C _Cmax , C _Dmax and two variables Y _C and T _C whose initial values are 0 are prepared in advance, and the maximum value C _max is calculated every time the maximum value C _max is obtained in each region. _max is C
_In the case of _xmax , the variable T _C is incremented, and in the case of C _ymax , the variable Y _C is incremented. When the increment processing is completed for all areas, the values of the variables Y _C and T _C are _{called and} the frequencies of the maximum values C _xmax and C _ymax are compared. That is, when the variable Y _C is larger, it is determined that the direction of characters in the document image 1 is horizontal writing. If the variable T _C is larger, vertical writing is determined.

In this embodiment, since the processing is performed by the one-dimensional Fourier transform, the number of times of processing per region is m times in the vertical direction and n times in the horizontal direction. On the other hand, in order to process the same area by the conventional two-dimensional Fourier transform, the number of times of processing the product of the orders m and n is necessary. That is, in this embodiment, (m + n) times,
In the conventional method, the number of times of Fourier transform processing is (m × n), and the ratio of both is (m + n) / (m × n).
The processing amount of the embodiment is reduced by an order of magnitude as compared with the conventional method.
As described above, in the embodiment, the calculation time is shortened to provide a practical processing speed, and it is possible to sufficiently withstand the use of OCR or the like.

In this embodiment, in order to deal with various document images, the document image 1 is divided into a plurality of regions, and the horizontal and vertical projection distributions of each region are Fourier transformed,
It is determined which axial direction the maximum value of the obtained harmonic component (power spectrum) is, and then the frequency of the maximum value for each axial direction is compared in the entire document image 1 to determine the document direction. ing. However, if the input is a uniform document image, the document image 1 can be processed as it is without being divided. In that case, the processing time is reduced almost in inverse proportion to the number of divisions.

[0014]

As described above, according to the present invention, the projection distributions in the X-axis and Y-axis directions of a document image are each subjected to a one-dimensional Fourier transform, and the values of the obtained power spectra are compared with each other to obtain a document It is determined whether the image is in vertical writing or horizontal writing. As a result, the processing time is greatly shortened as compared with the case of the conventional two-dimensional Fourier transform processing, and it can withstand practical use sufficiently, and further, the processing speed of the OCR connected in the subsequent stage can be improved. To do.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a document image.

FIG. 2 is an explanatory diagram showing a divided region and its projection distribution.

[Explanation of symbols]

 1 document image 2 characters A to D area

Claims (1)

[Claims] 1. A means for converting an input image signal into a binarized image, a means for obtaining a projection distribution in the X-axis and Y-axis directions of the binarized image, and a discrete value system 1 for each projection distribution. A means for obtaining a power spectrum by performing a four-dimensional Fourier transform and a means for comparing the values of each power spectrum and determining which direction of the input image is vertical writing or horizontal writing based on the magnitude relationship are provided. A device for determining the vertical and horizontal directions of a document.