Image analysis – Image transformation or preprocessing – Measuring image properties
Reexamination Certificate
1998-12-11
2001-01-16
Bella, Matthew C. (Department: 2721)
Image analysis
Image transformation or preprocessing
Measuring image properties
C382S174000, C382S194000, C382S289000
Reexamination Certificate
active
06175664
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a character reader for entering characters into a word processor or the like, and particularly to an optical character reader (hereinafter, referred to as an OCR) for reading and entering characters written on paper or the like by optical methods.
2. Description of the Related Art
As a means of entering characters into a word processor or the like, an OCR is used in some cases to read and enter the characters written on paper or the like automatically, without typing with a keyboard. Input of the characters by means of an OCR involves the steps of reading the characters on paper or the like with a scanner to convert them into image data, analyzing the layout of the image data to discern the character portion and recognizing the characters by the technique of pattern recognition. In entering the characters by means of an OCR, however, if the direction of the image input portion of a scanner tilts on the surface of the paper having the characters written thereon at the time of reading the characters with a scanner, thereby to cause the tilt in the readout image. Accordingly, there have been such problems that the characters can't be correctly discerned by the layout analysis and the performance of recognition is decreased because of recognition processing being performed with the characters tilted.
As countermeasures of these problems, there have been various conventional techniques to improve the recognition performance by detecting the characters tilting on the paper and correcting the detected tilt. In this kind of the conventional technique for use in an OCR, for example, “A Character String Direction Discrimination Device” (Article 1) is disclosed in Japanese Patent Laid-Open No. 61-160180. A character reader described in the same patent, comprises a photoelectric converter for converting characters from analog data to digital data by photoelectric methods so as to deliver the quantum signals, an image data storing unit for storing the delivered quantum signals as image data, a marginal distribution creating unit for requiring a histogram obtained by performing projection as for one region or more within the stored image from several directions and accumulating the density, and a character direction judging unit for judging the direction of the character string on the basis of the created histogram. The character reader requires a histogram of black pixel by the projection performed on the character string from several directions and finds the sharpest portion in the directions, which is recognized as the tilt of the above character string.
As another conventional technique, “A Character Reader” (Article 2) is disclosed in Japanese Unexamined Patent Publication (Kokai) No. Heisei 2-116987. The character reader described in the same patent comprises an input means for entering image data, an extracting means for extracting character string from the entered image data, a character discerning means for discerning each character from the extracted character string, and a reference line extracting means in which assuming a certain straight line passing a specified position like the lower end portion of a circumscribed rectangle of each character having been discerned, a histogram on the parameter space is required as for a set of parameter defining this straight line and the tilt on the straight line which is defined by a set of parameter providing with the maximum frequency on the histogram is regarded as the tilt on the character string. The character reader once scans the whole image data to be entered, requires a histogram on a parameter space by assuming the above specified straight line, after roughly discerning the characters, and recognizes the tilt on the character string on the basis of the required histogram.
As described above, the conventional character reader has a drawback that it takes much time in processing because the processing amount becomes huge in case of recognizing the tilt on the character string in order to improve accuracy of character recognition.
More specifically, the conventional character reader described in the article
1
requires a histogram of black pixel by the projection performed from several directions for all the black pixels in order to improve accuracy, with the result that processing amount becomes huge.
Additionally, since the processing amount for requiring a set of parameter defining a straight line which passes a specified position of a circumscribed rectangle of each character, is proportional to the number of characters, processing amount becomes enormous when there are many characters.
SUMMARY OF THE INVENTION
A first object of the present invention is to provide a character reader capable of improving accuracy of character recognition by recognizing the tilt of image data and correcting the same so as to perform character recognition, without requiring a histogram of black pixel by the projection performed from several directions as for all the black pixels, discerning the characters from the image data and scanning the whole image data in order to require the tilt.
A second object of the present invention is, in addition to the first object, to provide a character reader capable of decreasing the processing amount for detecting and correcting the tilt of the image data so to improve the processing speed, by removing the steps of requiring a histogram of black pixel by the projection from several directions as for all the black pixels in order to detect the tilt on the image data, discerning the characters from the image data and scanning the whole image data in order to require the tilt.
According to one aspect of the invention, a character reader for reading and entering characters by optical methods, comprises:
an input means for photoelectrically reading characters as an image, creating image data formed by black pixels lying on a white surface, and entering the image data;
a tangent detecting means for generating at least two line segments, each line segment connecting two black pixels within a character string of the image data; and
a tilt deciding means for choosing a representative line segment from the at least two line segments which is representative of the tilt of the character string, on the basis that the distance between two black pixels specifying the representative line segment is longer than the distance between two black pixels specifying another line segment.
In the preferred construction, the tangent detecting means may define a black pixel first detected by scanning the image data from the top portion by one line in a constant direction as a first black pixel, so to detect a first line segment extending from the first black pixel on the opposite side to the scanning direction, and defines a black pixel at the distal end on the opposite side to a group of black pixels sequentially lying from the first black pixel in the scanning direction as a second black pixel, so to detect a second line segment extending from the second black pixel on the same side as the scanning direction; and the tilt deciding means chooses the longer one of the first and second line segments as the representative line segment.
In another preferred construction, the tangent detecting means includes a reference point detecting means for detecting a first black pixel that is a black pixel first detected by scanning the image data from the top portion by one line in a constant direction and a second black pixel that is a black pixel located at the distal end on the opposite side to a group of black pixels sequentially lying from the first black pixel in the scanning direction, a first quasi-reference detecting means for detecting a third black pixel that is a black pixel first detected by scanning a first region obtained by a predetermined condition on the basis of the coordinates of the first black pixel, from the top portion by one line in the same direction as the scanning direction, and a second quasi-reference detecting means for detecting
Bella Matthew C.
Chawan Sheela
NEC Corporation
Scully Scott Murphy & Presser
LandOfFree
Optical character reader with tangent detection for... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Optical character reader with tangent detection for..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Optical character reader with tangent detection for... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2467414