Pattern features extracting apparatus and method

• US 4,543,660 A
• Filed: 04/14/1983
• Issued: 09/24/1985
• Est. Priority Date: 04/15/1982
• Status: Expired due to Fees

• Current Assignee: Tokyo Shibaura Denki Kabushiki Kaisha
• Original Assignee: Tokyo Shibaura Denki Kabushiki Kaisha
• Inventors: Maeda, Ken-ichi
• Primary Examiner(s): Boudreau, Leo H.
• Application Number:
  US06/485,061
• Time in Patent Office: 894 Days
• Field of Search: 382/25-28, 382/41-43, 382/36, 382/30, 382/34
• US Class Current: 382/218
• CPC Class Codes: G06V 10/20 Image preprocessing
• CPC Class Codes: G06V 10/42 Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
• CPC Class Codes: G06V 10/75 Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries

Abstract

A plurality of different mask pattern data which are indicated by values obtained from Hermitte'"'"'s polynomials of different degrees weighted by a Gaussian function are stored in mask memories, respectively. Each mask pattern data stored in each of the mask memories is convolved by a sum-of-product circuit together with unknown input pattern data. A plurality of extracted pattern feature data are stored in pattern feature memories.

Claims

1. A pattern features extracting apparatus comprising:

a plurality of mask memories for storing a plurality of different mask pattern data whose values are obtained from Hermite'"'"'s polynomials weighted by a Gaussian function;

an input pattern memory for storing unknown input pattern data;

means for accessing data in a predetermined area of said input pattern memory including a given position of the unknown input pattern; and

means for convolving the data accessed corresponding to the unknown input pattern data using the plurality of different mask pattern data in said plurality of mask memories.

2. An apparatus according to claim 1, wherein the mask pattern data stored in each of said plurality of mask memories are obtained from Hermite'"'"'s polynomials of different degrees weighted by the Gaussian function.

3. An apparatus according to claim 1, wherein the plurality of different mask pattern data include a first group of mask pattern data plotted in a first system of coordinates and a second group of mask pattern data plotted in a second system of coordinates which is rotated with respect to said first system through a predetermined angle.

4. An apparatus according to any one of claims 1 to 3, wherein said convolving means includes multiplying means for multiplying the unknown input pattern data read out from said input pattern memory with the mask pattern data read out from a given mask memory among said plurality of mask memories;

adding means for receiving output data from said multiplying means;

an accumulator for storing output data from said adding means; and

means for supplying output data from said accumulator together with the output data from said multiplying means to said adding means.

5. An apparatus according to claim 1, wherein said convolving means produces output data which is then stored in a buffer memory in a matrix form, so as to correspond to feature data and position data of a given position of the unknown input pattern, one of the feature and position data being defined as row data, and the other thereof being defined as column data.

6. A Pattern recognition system comprising:

a plurality of mask memories for storing a plurality of different mask pattern data whose values are obtained from Hermite'"'"'s polynomials weighted by a Gaussian function;

an input means for inputting unknown input pattern data;

means for convolving the mask pattern data and unknown input pattern data;

means for storing output data from said convolving means in a matrix form;

means for calculating a multiple similarity for each output data stored in said matrix storing means; and

means for recognizing the unknown input pattern in accordance with the multiple similarity.

7. A system according to claim 6, wherein the mask pattern data stored in each of said plurality of mask memories are obtained from Hermite'"'"'s polynomials of different degrees weighted by the Gaussian function.

8. A system according to claim 6, wherein the plurality of different mask pattern data include a first group of mask pattern data plotted in a first system of coordinates and a second group of mask pattern data plotted in a second system of coordinates which is rotated with respect to said first system through a predetermined angle.

9. A system according to any one of claims 6 to 8, wherein said convolving means includes multiplying means for multiplying the unknown input pattern data read out from said input means with the mask pattern data read out from a given mask memory among said plurality of mask memories;

adding means for receiving output data from said multiplying means;

an accumulator for storing output data from said adding means; and

means for supplying output data from said accumulator together with the output data from said multiplying means to said adding means.

10. A system according to claim 6, wherein said convolving means produces output data which is then stored in a buffer memory in a matrix form, so as to correspond to feature data and position data of a given position of the unknown input pattern, one of the feature and position data being defined as row data, and the other thereof being defined as column data.

11. A system according to claim 6, wherein said multiple similarity calculating means includes a plurality of multiple similarity calculating circuits for obtaining a plurality of multiple similarities between the unknown input pattern data and said output data stored in said matrix storing means for the unknown input pattern, and a weighted-sum circuit for summing said plurality of multiple similarities.