Apparatus and method for characterizing digital images

• US 20040008892A1
• Filed: 07/12/2002
• Published: 01/15/2004
• Est. Priority Date: 07/12/2002
• Status: Active Grant

• Current Assignee: CACI MTL Systems, Inc. (CACI International Incorporated)
• Original Assignee: MTL Systems, Inc.
• Inventors: Hirsch, Herbert L., Hill, Richard K.
• Granted Patent: US 7,233,696 B2
• Time in Patent Office: Days
• US Class Current: 382/228
• CPC Class Codes: G06V 10/42 Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation

Abstract

An automatic digital image characterization system has a feature extractor, including a segment processor and a feature processor. The segment processor is connected for receiving an image in the form of digitized pixel values; each pixel value having an amplitude and being associated with positional information in the form of column and row values. The feature processor converts the image information into column and row axis functions having calculated values of statistical mean amplitude and standard deviation. A system processor registers images, senses image changes, locates objects and detects hidden information.

Claims

1. A method of characterizing an image, comprising the steps of:

(1) generating a stream of digital information corresponding to a series of variations in said image;

(2) organizing said stream of digital information into segment sets delimited by local maxima/minima of said variations;

(3) generating a first axis function comprising mean values of digital information incorporated into said segments;

(4) generating a second axis function comprising standard deviations of digital information incorporated into said segments; and

(5) using said first axis function and said second axis function as characterization measures for said image.

2. A method of characterizing an image according to claim 1, wherein said segment sets comprise segment amplitude sets for M columns of said image, segment period sets for M columns of said image, segment amplitude sets for N rows of said image and segment period sets for N rows of said image.

3. A method of registering a second image with a first image comprising the steps of;

(1) generating a first stream of digital information corresponding to a first series of variations in said first image;

(2) generating a second stream of digital information corresponding to a second series of variations in said second image;

(3) organizing said first stream of information into first segment sets delimited by local maxima/minima of said first series of variations in said first image;

(4) organizing said second stream of information into second segment sets delimited by local maxima/minima of said second series of variations in said second image;

(5) generating a first-axis function comprising mean values of digital information incorporated into said first segment sets;

(6) generating a second axis function comprising mean values of digital information incorporated into said second segment sets;

(7) generating a third axis function comprising standard deviations for the mean values of digital information incorporated into said first segment sets;

(8) generating a fourth axis function comprising standard deviations for the mean values of digital information incorporated into said second segment sets;

(9) generating a difference coefficient by comparing values of said first, second, third and fourth axis functions; and

(10) using said difference-coefficient as a measure of accuracy for said registering.

4. Apparatus for characterizing a digitized, two-dimensional, column/row image, said apparatus comprising:

(a) a segment processor for segmenting said image into

(1) sets of image amplitude data for pixels arranged along a column axis,

(2) sets of image period data for pixels arranged along a column axis,

(3) sets of image amplitude data for pixels arranged along a row axis, and

(4) sets of image period data for pixels arranged along a row axis; and

(b) a feature processor for calculating statistical properties of image data, segmented as aforesaid by said segment processor.

5. Apparatus according to claim 4, said feature processor comprising means for calculating mean values and standard deviations of said image data.

6. An automatic image characterization system comprising:

(a) means for calculating image segment parameters;

(b) means for calculating image axis functions based upon said image segment parameters;

(c) means for determining registration of a reference image based upon said image axis functions; and

(d) means for locating a target image based upon said registration of said reference image.

7. The method of calculating column and row segment amplitudes and periods by treating each column and row of a two-dimensional array of image pixels as a signal-like data series.

8. The method of calculating image column and row axis functions, comprising the steps of:

(1) calculating the statistical means (averages) for a series of column segment amplitude sets

(2) organizing said statistical means into a series as the column segment amplitude mean axis function;

(3) calculating the statistical standard deviations for said column segment amplitude sets

(4) organizing said standard deviations into a series as the column amplitude deviation axis function;

(5) calculating the statistical means (averages) for said column segment period sets

(6) organizing said statistical means into a series as the column period mean axis function;

(7) calculating the statistical standard deviations for said column segment period sets

(8) organizing said statistical standard deviations into a series as the column period deviation axis function;

(9) calculating the statistical means (averages) for a series of row segment amplitude sets;

(10) organizing said statistical means into a series as the row amplitude mean axis function;

(11) calculating the statistical standard deviations for said row segment amplitude sets;

(12) organizing said standard deviations into a series as the row amplitude deviation axis function; and

(13) calculating the statistical means (averages) for said row segment period sets

(14) organizing said statistical means into a series as the row period mean axis function;

(15) calculating the statistical standard deviations for said column row segment period sets; and

(16) organizing said statistical standard deviations into a series as the row period deviation axis function.

9. The method of characterizing an M-column by N-row array of image pixels, said method comprising the steps of:

(1) reading the amplitudes of said pixels;

(2) assigning column and row indices to said amplitudes;

(3) organizing said amplitudes into segment amplitude sets for M columns and N rows;

(4) organizing segment period sets for M columns and N rows

(5) calculating mean value sets for said segment amplitude sets;

(6) calculating standard deviation sets for said mean value sets; and

(7) using said mean value sets and said standard deviation sets to characterize said array of image pixels.

10. The method of claim 9 wherein said mean value sets and said standard deviation sets are used to create difference integral curves, and said difference integral curves are used for calculating image registration errors.

11. The method of claim 9 wherein registration errors are determined through calculation of difference coefficients D(c) and D(r) defined by the following equations:

D(c)=[Ma(c,A)−

Ma(c,B)][Da(c,A)−

Da(c,B)][Mt(c,A)−

Mt(c,B)][Dt(c,A)−

Dt(c,B)]D(r)=[Ma(r,A)−

Ma(r,B)][Da(r,A)−

Da(r,B)][Mt(r,A)−

Mt(r,B)][Dt(r,A)−

Dt(r,B)]

where;

Ma(c,X)=Column amplitude mean axis function for image X, where “

c”

denotes a particular column

Da(c,X)=Column amplitude deviation axis function for image X, where “

c”

denotes a particular column

Mt(c,X)=Column period mean axis function for image X, where “

c”

denotes a particular column

Dt(c,X)=Column period deviation axis function for image X, where “

c”

denotes a particular column

Ma(r,X)=Row amplitude mean axis function for image X, where “

r”

denotes a particular row,

Da(r,X)=Row amplitude deviation axis function for image X, where “

r”

denotes a particular row,

Mt(r,X)=Row period mean axis function for image X, where “

r”

denotes a particular row,

Dt(r,X)=Row period deviation axis function for image X, where “

r”

denotes a particular row.

12. The method of claim 11 wherein the quantities in at least one of said bracketed expressions are converted to absolute values.

13. The method of claim 11 wherein the Image A and Image B quantities of at least one of said axis functions are interchanged.

14. The method of claim 11 wherein at least one of said bracketed quantities has a value of zero.

15. The method of claim 11 wherein at least one of said axis functions is replaced by a normalized version thereof.

16. The method of locating objects from a given image in other images by using a measure of a difference in an axis function.

17. The method according to claim 16 wherein said measure of a difference is an aggregate difference coefficient established without use of row data.

18. The method according to claim 16 wherein said measure of a difference is an aggregate difference coefficient established without use of column data.

19. The method according to claim 16 wherein said measure of a difference is an aggregate difference coefficient established through the use of product data.

20. The method according to claim 16 wherein said measure of a difference is an aggregate difference coefficient established through use of a weighted sum.

21. The method of determining the presence of hidden information in an image by using measures of differences in segment amplitude and segment period data sets.

22. The method of characterizing an image comprising the steps of:

(1) generating a series of digital codes representing samples of image taken along a series of points therein,

(2) calculating a plurality of mean values of said codes,

(3) calculating standard deviations statistically relating said codes and said mean values, and

(4) establishing an image characterization parameter based upon said mean values and said standard deviations.

23. An image characterization system comprising:

(a) An image source supplying M columns and N rows of digital information representing an image to be characterized,

(b) a Segment Processor coupled to said image source for reorganizing said digital information into segment amplitude sets and segment period sets for said M columns and said N rows, and

(c) a feature processor for characterizing said image by calculating mean values and standard deviations of data comprising said segment sets.

24. Apparatus according to claim 23, further comprising a second said image source, a second said segment processor, a second said feature processor and means for calculating a coefficient indicating a difference between image data supplied by said image source and image data supplied by said second image source.

25. A method for identifying a change of environment comprising the steps of:

providing a first image of said environment;

providing a second image of said environment;

generating first digital information corresponding to a series of variations in said first image;

generating second digital information corresponding to a series of variations in said second image;

representing said first and second digital information with a plurality of functions; and

using said plurality of functions to perform a comparison of said first image to said second image to detect said change of environment.

26. The method as recited in claim 25 wherein said method further comprises the steps of:

organizing said first digital information into first segment sets delimited by local maxima/minima of said first series of variations in said first image;

organizing said second digital information into second segment sets delimited by local maxima/minima of said second series of variations in said second image.

27. The method as recited in claim 26 wherein said method further comprises the steps of:

generating first axis function comprising mean values of digital information incorporated into said first segment sets; and

generating second axis function comprising mean values of digital information incorporated into said second segment sets.

28. The method as recited in claim 27 wherein said method further comprises the steps of:

generating a third axis function comprising standard deviations for the mean values of digital information incorporated into said first segment sets; and

generating a fourth axis function comprising standard deviations for the mean values of digital information incorporated into said second segment sets.

29. The method as recited in claim 27 wherein said method further comprises the steps of:

generating a third axis function comprising standard deviations for the mean values of digital information incorporated into said first segment sets; and

generating a fourth axis function comprising standard deviations for the mean values of digital information incorporated into said second segment sets.

30. The method as recited in claim 27 wherein said method further comprises the steps of:

generating a difference coefficient by comparing values of said first and second axis functions.

31. The method as recited in claim 29 wherein said method further comprises the steps of:

generating a difference coefficient by comparing values of said first, second, third and fourth axis functions.

32. The method as recited in claim 31 wherein said using step further comprises the step of:

using said difference coefficient as a measure of accuracy for said registering.

41. The method as recited in claim 32 wherein said method further comprises the steps of:

generating at least one function for each of said first and second images; and

using each of said at least one function to determine variations between said first and second images.

33. A method for characterizing an image, said method comprising the steps of:

providing first image pixel data for a first image;

generating a first signal-like data series for said first image pixel data; and

determining row and column segment amplitudes and periods for said first image using said first image data.

34. The method as recited in claim 33 wherein said method further comprises the steps of:

providing second image pixel data for a second image;

generating a second signal-like data series for said second image pixel data; and

determining row and column segment amplitudes and periods for said second image using said second image data.

35. The method as recited in claim 34 wherein said method further comprises the steps of:

using said first and second signal-like data series to perform a predetermined task.

36. The method as recited in claim 35 wherein said predetermined task comprises the step of:

determining an image registration.

37. The method as recited in claim 35 wherein said predetermined task comprises the step of:

determining changes between said first and second images.

38. The method as recited in claim 35 wherein said predetermined task comprises the step of:

locating an object in at least one of said first or second images.

39. The method as recited in claim 35 wherein said predetermined task comprises the step of:

locating an object in each of said first or second images.

40. The method as recited in claim 35 wherein said predetermined task comprises the step of:

determining a presence or absence of hidden information in at least one of said first or second images.

42. The method as recited in claim 39 wherein said method further comprises the step of:

generating at least one axis function for each of said first and second images.

43. The method as recited in claim 39 wherein said method further comprises the step of:

generating at least one segment function for each of said first and second images.

44. The method as recited in claim 42 wherein said method further comprises the step of:

generating at least one segment function for each of said first and second images.

45. The method as recited in claim 42 wherein said method further comprises the steps of:

generating rows and columns of pixel data for each of said first and second images; and

determining statistics to provide image properties for said first and second images.

46. The method as recited in claim 45 wherein said method further comprises the steps of:

determining amplitude and period means and standard deviations to provide said period statistics; and

using said amplitude and period means and standard deviations to provide said image properties.

47. The method as recited in claim 46 wherein said image properties comprises at least one of the following:

registration, changes between said first and second images, presence or absence of hidden information, or locating an object in at least one of said images.

48. An automatic image characterization system comprising:

(a) means for calculating image segment parameters;

(b) means for calculating image axis functions;

(c) first means for determining image registration;

(d) second means for determining changes between two images;

(e) third means for locating an object from one image in another image; and

(f) fourth means for determining the presence or absence of hidden information in images.

49. The method of treating each row and column of image pixels as a signal-like data series, for the purpose of calculating row and column segment amplitudes and periods.

50. The method of calculating image row and column axis functions, comprising the steps of:

(a) calculating the statistical mean (average) for a column segment amplitude set and organizing these data into a series as the column amplitude mean axis function;

(b) calculating the statistical standard deviation for each column segment amplitude set and organizing these data into a series as the column amplitude deviation axis function;

(c) calculating the statistical mean (average) for each column segment period set and organizing these data into a series as the column period mean axis function;

(d) calculating the statistical standard deviation for each column segment period set and organizing these data into a series as the column period deviation axis function;

(e) calculating the statistical mean (average) for each row segment amplitude set and organizing these data into a series as the row amplitude mean axis function;

(f) calculating the statistical standard deviation for each row segment amplitude set and organizing these data into a series as the row amplitude deviation axis function;

(g) calculating the statistical mean (average) for each row segment period set and organizing these data into a series as the row period mean axis function; and

(h) calculating the statistical standard deviation for each row segment period set and organizing these data into a series as the row period deviation axis function.

51. The method of using measures of differences in axis functions for the purpose of calculating image registration errors.

52. The method of using measures of differences in axis functions for the purpose of determining changes between images.

53. The method of using measures of differences in axis functions for the purpose of locating objects from a given image in other images.

54. The method of using measures of differences in segment amplitude and segment period data sets for the purpose of determining whether or not an image contains hidden information.