Image recognition system

• US 5,033,099 A
• Filed: 07/30/1990
• Issued: 07/16/1991
• Est. Priority Date: 07/31/1989
• Status: Expired due to Fees

• Current Assignee: Agency for Industrial Science and Technology
• Original Assignee: Agency for Industrial Science and Technology
• Inventors: Yamada, Hiromitsu, Saito, Taiichi, Yamamoto, Kazuhiko
• Primary Examiner(s): BRINICH, STEPHEN M
• Application Number:
  US07/559,588
• Time in Patent Office: 351 Days
• Field of Search: 382/21, 382/22
• US Class Current: 382/197
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/182 by coding the contour of the pattern
• CPC Class Codes: G06V 30/196 using sequential comparisons of the image signals with a plurality of references

Abstract

A model for a previously defined form is provided in terms of directional features for particular evaluation points and shift vectors from one evaluation point to the next. An input image is represented by density gradients for different directional planes. Shift operations for a counter plane which is used to accumulate values for evaluating the similarity between the image and the model, and add operations in which values from input directional planes as specified by the model are added to the counter plane, are performed repeatedly by parallel processing to find the total match with the model for the form.

Claims

1. An image recognition apparatus for recognizing an image by extracting a previously defined form from an input image composed of a plurality of image points each of which has image information, and to which predetermined multiple directions are assigned, said image recognition apparatus comprising:

model storing means for storing a model of said form, said model being expressed in terms of a sequence consisting of evaluation points defined on said form, each of said evaluation points having a value for a directional feature determined for the respective evaluation point and having a shift vector to said respective evaluation point from an evaluation point immediately preceding said respective evaluation point in said sequence;

directional plane generator means for generating a plurality of directional planes, each directional plane corresponding to a respective one of said multiple directions and including a plurality of directional plane points, each directional plane point corresponding to a respective one of said image points and having a respective directional intensity value which represents the change in the density of said input image in the one of said multiple directions that corresponds to the directional plane to which the respective directional plane point belongs;

directional plane storing means for storing said directional planes generated by said directional plane generator means;

counter plane storing means for storing a counter plane which includes a plurality of counter plane points each of which corresponds to a respective one of said image points, said counter plane points having counter values each of which represents a degree of similarity in form between said input image and said model;

shift means for shifting said counter plane stored in said counter plane storing means according to said shift vector defined for each evaluation point;

add means for adding, to said counter values of said counter plane shifted by said shift means, the directional intensity values of a directional plane determined by the value for the directional feature defined for the relevant evaluation point with regard to the shifting operation by said shift means;

control means for performing the shift operation by said shift means and the add operation by said add means repeatedly in the order of said evaluation points in said sequence; and

detect means for determining the location of said form from the distribution of said counter values for said counter plane points of said counter plane storing means, after said shift operation and said add operation have been performed repeatedly.

2. An image recognition apparatus as claimed in claim 1, wherein each of the directional intensity values which are added by said add means add to said counter values is a maximum out of a value of each of said directional plane points and values of points adjacent to each of said directional plane points.

3. An image recognition apparatus as claimed in claim 1, wherein said add means adds to said counter values, a maximum out of directional intensity values of a particular directional plane for a direction as specified by said value for the directional feature and directional intensity values of the directional planes for adjacent directions in said multiple directions to the direction of said particular directional plane.

4. An image recognition apparatus as claimed in claim 1, wherein said add means adds to the counter values each of which is a maximum out of a counter value of each of said counter plane points and counter values of points adjacent to each of said counter plane points, said directional intensity values of said directional plane.

5. An image recognition apparatus as claimed in claim 1, wherein said add means adds to said counter values, a maximum value out of directional intensity values of a particular directional plane for a direction as specified by said value for the directional feature and the directional intensity values of the directional planes for the directions opposed to that of said particular directional plane.

6. An image recognition method for recognizing an image by extracting a previously defined form from an input image composed of a plurality of image points each of which has image information, and to which predetermined multiple directions are assigned, said image recognition apparatus comprising the steps of:

storing a model of said form, said model being expressed in terms of a sequence consisting of evaluation points defined on said form, each of said evaluation points defined on said form, each of said evaluation points having a value for a directional feature determined for the respective evaluation point and having a shift vector to said respective evaluation point from an evaluation point immediately preceding said respective evaluation point in said sequence;

generating directional planes, each directional plane corresponding to a respective one of said multiple directions and including a plurality of directional plane points, each directional plane point corresponding to a respective one of said image points and having a respective directional intensity value which represents the change in the density of said input image in the one of said multiple directions that corresponds to the directional plane to which the respective directional plane point belongs;

storing said directional planes;

storing a counter plane which includes a plurality of counter plane points each of which corresponds to a respective one of said image points, said counter plane points having counter values each of which represents a degree of similarity in form between said input image and said model;

shifting said stored counter plane according to said shift vector defined for each evaluation point;

adding to said counter values of said counter plane shifted by said shift means, the directional intensity values of a directional plane determined by the value for the directional feature defined for the relevant evaluation point with regard to the shifting operation by said shift means;

performing the shift operation and the add operation repeatedly in the order of said evaluation points in said sequence; and

determining the location of said form from the distribution of said counter values for said stored counter plane points, after said shift operation and said add operation have been performed repeatedly.

7. An image recognition method as claimed in claim 6, wherein each of the directional intensity values which are added to said counter values is a maximum out of a value of each of said directional plane points and values of points adjacent to each of said directional plane points.

8. An image recognition method as claimed in claim 6, wherein a maximum out of directional intensity values of a particular directional plane for a direction as specified by said value for the directional feature and directional intensity values of the directional planes for adjacent directions in said multiple directions to the direction of said particular directional plane, is added to said counter value.

9. An image recognition method as claimed in claim 6, wherein said directional intensity values of said directional plane are added to the counter values each of which is a maximum out of a counter value of each of said counter plane points and counter values of points adjacent to each of said counter plane points.

10. An image recognition method as claimed in claim 6, wherein a maximum out of directional values of a particular directional plane for a direction as specified by said value for the directional feature and the directional intensity values of the directional planes for the directions opposed to that of said particular directional plane, is added to said counter value.