JPS59226978A - Skew correction system - Google Patents

Abstract

PURPOSE:To correct the tilt of a general document picture by obtaining the dispersion of the oblique shadow value on a scan line and using the scan number N which minimizes said shadow value to perform the 1/N scan of a picture image. CONSTITUTION:The dispersion sigma generated from the oblique shadow value on each scan line for a period from the start of appearance of a black picture element to disappearance of the black picue element or in a range up to several hundred lines is obtained by a sigma calculating circuit 5. Then the tilt of pictures is reduced as the dispersion sigma is reduced in terms of the same picture. Thus a skew detecting circuit 6 evaluates the obtained dispersion sigma to detect the tilting degree of pictures. Then the 1/N scan is performed via a controller 2 for pictures stored in a picture memory 1 by means of the scan number N which minimizes the dispersion sigma in accordance with the result of detection of the circuit 6. As a result, the skew corrected picture signals with which no tilt is produced with pictures are delivered successively from a memory 1.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field The present invention relates to a skew correction method for correcting the skew of an image read by a scanner.

PRIOR ART In general, in OCR, if the image that is captured by the scanner and read is tilted, there is a high possibility that errors or digits will occur when pattern recognition is performed on cut-out characters9. Therefore, it is necessary to perform skew correction as a preprocessing for character extraction.

Conventionally, this type of skew correction method uses a special form that has marks or lines for tilt detection in advance, and detects the marks when it is read by a scanner to determine the degree of skew of the form. According to the detected inclination, arithmetic processing is performed to transform the position of the image, and data in the image memory is rewritten.

However, with these conventional skew correction methods, the skew cannot be detected without the use of a special form with marks or lines attached in advance, making it impossible to apply it to general documents. It is now impossible to correct the inclination of the image in any way.

Purpose The present invention has been made in consideration of the above points, and provides a skew correction force formula that can correct the skew of general document images.

composition An embodiment of the present invention will be described in detail below.

In the skew correction force formula according to the present invention, by determining the variance σ of the shading value (count value of the number of black pixels) on the set meal line by the scanner, the skew value is determined at around 1 of the image, and the skew value The skew correction is performed by using the number N of set meals that minimizes the variance σ of 1/N of the image memory in which data of images read by the scanner is stored.

FIG. 1 shows an example of a system configuration for concretely implementing the skew correction force formula according to the present invention. An image memory 1 that sequentially stores converted input image signals and its memory controller 2. A counter that counts the number of black pixels on each scanning line of the input image signal to obtain a shading value. 71: Stores the shading values on the scan line (shading memory 4. Calculate the variance σ of the shading values from the memory contents) σ
Calculation circuit 5. A skew detection circuit 57) h performs skew detection using the obtained variance σ as an evaluation value and outputs a scanning command to the memory controller 2.

With such a configuration, in the present invention, the σ calculation circuit 5 calculates the variance σ based on the oblique shadow value on each scanning line from the time a black pixel begins to appear until the black pixel ceases to exist or up to the 100-odd line. , for the same image, the smaller the variance σ, the smaller the image tilt. Therefore, the skew detection circuit 6 detects the degree of image tilt by evaluating the calculated variance σ, and depending on the detection result, The image stored in the image memory 1 is subjected to l/N scanning (diagnosis in FIG. 2) via the memory controller 2 using 1-N to minimize the variance σ of the oblique values. Then, by performing the 1/N regular scan, the image memo IJ and skew-corrected image signals with no image inclination are sequentially output and sent to an OCR character cutting device or the like.

At this time, if the number of scans N is changed by a minute amount, it will take time to process, so if you want to correct the tilt of the image within the allowable limit, use that range as a reference
The amount of change may be determined appropriately. For example, if the line spacing is 3 inches for inch lO type characters, the allowable slant range is 1.4° (N key 41), and when the line pitch is 6 inches, the allowable slant range is 0.2° (N = 28
6).

Now, a 1/N ffi scan of the image memory l in the case where the permissible tilt range is set to 0.2 degrees will be described.

First, the image memory 1 is read out and scanned at +0.2°, that is, 1/286, and based on the image signals on each read scanning line, the counter 3° oblique shadow memory 4 and the σ calculation circuit 5 calculate the oblique value. The variance σ2 is determined, and the skew detection circuit 6 compares it with the variance σ1 of the shading value previously determined when writing the image signal to the image memory 1, thereby making the following determination.

When σ1≧σ2, DIR="1" (positive direction) σ1<σ
2, DIR=“0” (negative direction), then I) rR=
If it is determined to be "l", the skew detection circuit 6 sends the image memo IJl through the memory controller 2 to sequentially scan 1/N of the image memo IJl tilted in the positive direction [0.4°→N=143,
0.6°→N=95. ...], σn-1 × σ. When , the previous N is adopted. Also, DIR=
If the skew detection circuit 6 is determined to be "0", the skew detection circuit 6 sequentially scans the image memo 'Jl by 1/N (-0, 2 degrees) tilted in the negative direction via the memory controller 2.

-〇, 4°, -0.6°,...], and σ. −
5〈σ. When , the previous N is adopted.

By performing 1/N i of image memory 1 while changing N within the image tilt tolerance range in this way,
It becomes possible to perform efficient skew correction with a reduction in processing time.

-Also, when performing 1/N scanning of the image memory l while changing N within the image inclination tolerance range, input image signals are stored in multiple image memories respectively, and the variance σ of each oblique value is If they are calculated in parallel at the same time, it becomes possible to perform the processing at higher speed.

Now, for example, when the line spacing is 6 inches for a type character of 10 inches, the number of times the image memory is scanned is 2 as described above. As shown in Fig. 1', image memories corresponding to the number of scanned images can be prepared to perform parallel processing. In other words, record the input image signal for each image! J 11, 1
At the same time as accumulating each of the 24 items, the counter 31
, oblique shadow memory 41. In the if system consisting of 51 σ calculating circuits, the variance σ of the oblique shadow value is calculated when the image inclination tolerance range is 0°, and the second and third systems calculate the distribution σ. ll valley image memo IJ
11 and 12 are scanned using their respective N values, the variance σ of each oblique value is obtained in parallel, and the skew detection circuit 6 selects N that provides the minimum elongation.

As described above, the skew correction method according to the present invention includes a means for determining the variance σ of the shading values on each scanning line of a binary image read by a scanner, and a means for determining the number of scans that minimizes the variance σ of the shading values. and means for performing 1/N scanning of the image memory in which the data of the binary image read by the scanner is stored, thereby correcting the inclination of the binary image read by the scanner. However, since the image processing means is used to determine the degree of image inclination, instead of detecting the inclination of the image using marks placed on the form as in the past, ordinary documents without any marks or the like are used. This method has an excellent advantage of being able to reliably perform skew correction on document images.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an example of a system configuration for concretely implementing the skew correction method according to the present invention, FIG. 2 is a diagram showing a 1/N scanning state of the image memory, and FIG. It is a block diagram showing another system configuration example for concretely implementing the invention. 1, 11, 12... image memory 2, 21,
22...Memory controller 3, 31, 32
, 33... Counter 4, 41, 42, 4
3...-oblique shadow memory 5, 51 +52, 53
...σ Calculation Circuit 6...Skew Detection Circuit Applicant's Agent Masaaki Umai Figure 1 Figure 2

Claims (1)

[Claims] Data of the binary image read by the scanner using means for calculating the variance σ of the shading values on each scanning line of the n binary images read by the scanner, and a set number that minimizes the variance σ of the shading values. A skew correction method that corrects the inclination of a binary image read by a scanner by performing a 1/N scan of an image matrix in which 1/N is stored.