Image recognition method and image recognition apparatus

• US 20050201622A1
• Filed: 03/11/2005
• Published: 09/15/2005
• Est. Priority Date: 03/12/2004
• Status: Active Grant

• Current Assignee: Panasonic Corporation (Panasonic Holdings Corporation)
• Original Assignee: Panasonic Corporation (Panasonic Holdings Corporation)
• Inventors: Takarada, Shinichi
• Granted Patent: US 7,751,610 B2
• Time in Patent Office: Days
• US Class Current: 382/218
• CPC Class Codes: G06V 10/758 Involving statistics of pixels or of feature values, e.g. histogram matching

Abstract

Provided is an image recognition method in which a first pattern area which is set inside a target to be recognized and a second pattern area which is set outside the target is used as a mask pattern, and when the comparison circuit judges that a difference obtained by subtracting by a differential circuit the largest luminance value in the target image contained in the first pattern area, which is detected by the largest value detection circuit, from the smallest luminance value in the target image contained in the second pattern area, which is detected by the smallest value detection circuit, is larger than a certain offset amount, the matching judgement that a predetermined pattern is present is made, and this recognition judgement is performed with the whole target image being scanned, thereby substantially reducing the processing load for the pattern matching.

Claims

1. An image recognition method in which a mask pattern indicating characteristics of image patterns to be recognized is used to recognize the image patterns from target image data having digitized gradations of two or more values, comprising:

said mask pattern having a first pattern area and a second pattern area;

obtaining a first representative value which is a luminance of the target image data from the first pattern area;

obtaining a second representative value which is a luminance of the target image data from the second pattern area; and

judging that an image to be recognized is present in the target image when a result of comparison between the first and the second representative values satisfies a predetermined condition.

2. The image recognition method as defined in claim 1, wherein

as the mask pattern to be used, the first pattern area in the mask pattern is positioned inside the second pattern area, and the first and second pattern areas are composed of a plurality of pixels.

3. The image recognition method as defined in claim 1, wherein

the first representative value is the largest value of the luminances of the target image data, which are obtained from the first pattern area, and the second representative value is the smallest value of the luminances of the target image data, which are included in the second pattern area.

4. The image recognition method as defined in claim 3, wherein

when the difference between the second representative value and the first representative value is equal to or larger than a predetermined threshold value as the predetermined condition, it is judged that the image to be recognized is present in the target image.

5. The image recognition method as defined in claim 1, wherein

the first representative value is an average value of the luminances of the target image data, which are obtained from the first pattern area, and the second representative value is an average value of the luminances of the target image data, which are obtained from the second pattern area.

6. The image recognition method as defined in claim 5, wherein

when the difference between the second representative value and the first representative value is equal to or larger than a predetermined threshold value as the predetermined condition, it is judged that the image to be recognized is present in the target image.

7. The image recognition method as defined in claim 1, wherein

the first representative value is a primary largest value and a primary average value of the luminances of the target image data, which are obtained from the first pattern area, and the second representative value is a secondary largest value and a secondary average value of the luminances of the target image data, which are obtained from the second pattern area.

8. The image recognition method as defined in claim 7, wherein

when the difference between the secondary largest value and the primary largest value is equal to or larger than the first predetermined threshold value and the difference between the secondary average value and the primary average value is equal to or larger than the second predetermined threshold value as the predetermined condition, it is judged that the image to be recognized is present in the target image.

9. The image recognition method as defined in claim 1, comprising:

enlarging or reducing the first pattern area and the second pattern area at the same ratio; and

sequentially evaluating the degrees of matching at the respective positions with the positions of the mask patterns being displaced relatively to the target image data, using the mask patterns of plural sizes.

10. The image recognition method as defined in claim 1, comprising:

turning the first pattern area and the second pattern area by the same angle; and

sequentially evaluating the degrees of matching at the respective positions with the positions of the mask patterns being displaced relatively to the target image data, using the mask patterns which are obtained by turning the mask pattern multiple times.

11. The image recognition method as defined in claim 1, comprising:

enlarging or reducing the target image data; and

sequentially evaluating the degrees of matching at the respective positions with the positions of the mask patterns being displaced relatively to the target image data, using the target image data and the mask patterns which have plural sizes.

12. The image recognition method as defined in claim 1, comprising:

turning the mask pattern; and

sequentially evaluating the degrees of matching at the respective positions with the relative positions between the mask patterns and the target image data being displaced, using the mask patterns at multiple turning angles which are obtained by turning the mask pattern.

13. The image recognition method as defined in claim 1, comprising:

a plurality of combinations of the first pattern area and the second pattern area; and

sequentially evaluating the degrees of matching at the respective positions with the positions of the mask patterns being displaced relatively to the target image data, using the plurality of mask patterns.

14. The image recognition method as defined in claim 1 comprising:

discriminating a position where a specific image pattern in the target image data is present and binarizing the target image data in the vicinity of the position.

15. The image recognition method as defined in claim 14, comprising:

a binarized area having a positional relationship which is fixed relatively to the first pattern area and the second pattern area; and

binarizing the target image data in the binarized area when the image pattern is judged as being present.

16. The image recognition method as defined in claim 15, comprising:

obtaining a threshold value on the basis of the representative value of the luminance of the target image data in the first pattern area and the representative value of the luminance of the target image data in the second pattern area; and

binarizing the target image data in the binarized area using the threshold value.

17. The image recognition method as defined in claim 16, wherein

an intermediate value between the representative value of the luminance of the target image data in the first pattern area and the representative value of the luminance of the target image data in the second pattern area is used as the threshold value.

18. The image recognition method as defined in claim 16, wherein

an intermediate value between the largest value of the luminances of the target image data, which are contained in the first pattern area, and the smallest value of the luminances of the target image data, which are contained in the second pattern area, is used as the threshold value.

19. The image recognition method as defined in claim 14, comprising:

a fixed value area having a positional relationship which is fixed relatively to the first pattern area, the second pattern area, and the binarized area; and

binarizing the target image data in the fixed value area so that the target image data have a predetermined one of two values when the image pattern is judged as being present.

20. The image recognition method as defined in claim 14, wherein

the portion which is not binarized is set to a predetermined one of the two values to obtain the binarization result of the whole image.

21. An image recognition apparatus which uses the image recognition method as defined in claim 1 to process an image, comprising:

a display unit for displaying the first pattern area and the second pattern area on a screen; and

a position specification unit for determining positions of the respective pixels in the first pattern area and the second pattern area on the screen.

22. The image recognition apparatus as defined in claim 21, comprising:

an edition unit for overlay-displaying the first pattern area and the second pattern area on the displayed image of the target image data to edit the first pattern area and the second pattern area when the first pattern area and the second pattern area are determined.

23. The image recognition apparatus as defined in claim 22, wherein

the target image data are displayed as luminance components, and the first pattern area and the second pattern area are displayed as color difference components different from each other.

24. The image recognition apparatus as defined in claim 21, comprising:

an edition unit for overlay-displaying the first pattern area, the second pattern area, and the binarized area on the displayed image of the target image data to edit the first pattern area, the second pattern area, and the binarized area, when the first pattern area, the second pattern area, and the binarized area are determined.

25. The image recognition apparatus as defined in claim 24, wherein

the target image data are displayed as luminance components, and the first pattern area, the second pattern area, and the binarized area are displayed as color difference components different from each other.

26. The image recognition apparatus as defined in claim 21, comprising:

an edition unit for overlay-displaying the first pattern area, the second pattern area, the binarized area, and the fixed value area on the displayed image of the target image data to edit the first pattern area, the second pattern area, the binarized area, and the fixed value area, when the first pattern area, the second pattern area, the binarized area, and the fixed value area are determined.

27. The image recognition apparatus as defined in claim 26, wherein

the target image data are displayed as luminance components, and the first pattern area, the second pattern area, the binarized area, and the fixed value area are displayed as color difference components different from each other.

28. The image recognition apparatus as defined in claim 21, wherein

the first pattern area and the second pattern area are recorded as relative positions with respect to a pixel to be noted.

29. The image recognition apparatus as defined in claim 21, wherein

the first pattern area, the second pattern area, and the binarized area are recorded as relative positions with respect to the pixel to be noted.

30. The image recognition apparatus as defined in claim 21, wherein

the first pattern area, the second pattern area, the binarized area, and the fixed value area are recorded as relative positions with respect to the pixel to be noted.

31. An image recognition apparatus which uses the image recognition method as defined in claim 4, comprising:

a unit for obtaining the predetermined threshold value by utilizing a luminance distribution of the target image.

32. The image recognition apparatus as defined in claim 31, comprising:

a unit for obtaining the predetermined threshold value by utilizing a luminance value which most frequently appears in the luminance distribution of the target image.