Image analysis – Histogram processing – For setting a threshold
Patent
1990-03-20
1993-03-23
Couso, Jose
Image analysis
Histogram processing
For setting a threshold
382 9, 382 48, G06K 962
Patent
active
051971070
DESCRIPTION:
BRIEF SUMMARY
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a character recognition apparatus. The present invention relates, in particular, to a character recognition apparatus which inputs image information including characters such as those in a document, as image data, by an image scanner or a facsimile, extracts a characteristic quantity of a character from the image data, compares the extracted characteristic quantity with standard characteristic quantities which are memorized in a dictionary, and recognizes the character.
2. Description of the Related Art
FIG. 1 shows a flow of main processes in the conventional character recognition apparatus.
In the step 401, the image of a document, characters which are to be recognized, is input as image data by an image scanner or the like, and is stored in a memory.
In the step 402, a region wherein successive character strings, i.e., a region wherein sentences are printed, is extracted and distinguished from other regions such as a picture, or a drawing, or the like.
In the step 403, a region of each character string is extracted from the above region wherein sentences are printed.
In the step 404, a region of each character (a character image) is extracted from the above character string.
In the step 406, a characteristic quantity of the character is extracted from the above modified character image by a predetermined procedure.
In the step 407, the above extracted characteristic quantity is compared with each of the plurality of characters which are memorized in the dictionary, one by one, and then a character which is most similar to the extracted characteristic quantity is recognized as a character which the above character image represents (step 488).
Next, in the step 489, it is determined whether or not the result of above character recognition is correct. If it is determined that the above recognition result is incorrect, a correction is carried out in the step 490, and the corrected character recognition result is stored in the memory in the step 491.
An example of the above procedure for extracting a characteristic quantity is shown in FIGS. 2 to 6.
Namely, when a character image as shown in FIG. 2 is extracted from an image which is input by an image scanner or the like, next, a contour of the character image is extracted as shown in FIG. 3, and the contour image is divided into a n.times.m (for example, 8.times.8) meshes. Then, the above contour line is decomposed into directional line segments in each mesh as shown in FIG. 4, and the number of directional line segments in each direction is detected in each mesh. FIG. 4 is a magnified view of the mesh (8, 1) of FIG. 3, where the mesh (8, 1) is divided into a further fine 8.times.8 meshes. Then, aspects of connection of contour points on the contour of the character image in the respective meshes are accumulated for four directions as shown in FIG. 5 to extract a characteristic quantity (A.sub.i,j, B.sub.i,j, D.sub.i,j).
Here, the above directions can be classified, for example, as shown in FIG. 5, into four directions: a right-left direction, a right-down direction, an up-down direction, and a left down direction, where the right-left direction is denoted by "0", the right-down direction is denoted by "1", the up-down direction is denoted by "2", and the left-down direction is denoted by "3". In the determination result of FIG. 4, A.sub.i,j, B.sub.i,j, C.sub.i,j, D.sub.i,j, respectively denote a number of directional line segments in the direction "0", a number of directional line segments in the direction "1", a number of directional line segments in the direction "2", and a number of directional line segments in the direction "3", in the mesh i,j. Thus, the characteristic quantity (A.sub.8,1, B.sub.8,1, C.sub.8,1, D.sub.8,1) of FIG. 4 is (11, 0, 2, 4). Namely, a four-dimensional vector quantity, which four dimensions correspond to the above four directions, is extracted in each mesh, and thus, four-dimensional vector quantities (A.sub.i,j, B.sub.i,j,C.sub.i,j, D.sub.i
REFERENCES:
patent: 4601054 (1986-07-01), Watari et al.
patent: 4606069 (1986-08-01), Johnsen
patent: 4641355 (1987-02-01), Hongo et al.
patent: 4701961 (1987-10-01), Hongo
patent: 4769851 (1988-09-01), Nishijima et al.
patent: 4837842 (1989-06-01), Holt
patent: 4887303 (1989-12-01), Hongo
patent: 4977603 (1990-12-01), Irie et al.
Edanami Takafumi
Fukuyama Noriyuki
Iwaki Hiroshi
Katsuyama Yutaka
Couso Jose
Fujitsu Limited
LandOfFree
Character recognition apparatus does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Character recognition apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Character recognition apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1357964