Method of and apparatus for analyzing images and deriving binary image representations

• US 5,848,198 A
• Filed: 01/12/1995
• Issued: 12/08/1998
• Est. Priority Date: 10/08/1993
• Status: Expired due to Fees

• Current Assignee: Alan Penn & Associates, Inc.
• Original Assignee: Alan Irvin Penn
• Inventors: Penn, Alan Irvin
• Primary Examiner(s): Couso, Jose L.
• Application Number:
  US08/371,918
• Time in Patent Office: 1,426 Days
• Field of Search: 382/276, 382/278, 382/295, 382/296, 382/298, 382/232, 382/248, 382/249
• US Class Current: 382/276
• CPC Class Codes: G06T 7/0012 Biomedical image inspection

Abstract

Data from multiple image sources are processed with a memory storing a library of data catalogs, each corresponding to one of the image sources. The catalogs have entries corresponding to the images. Each catalog entry has signal values specifying how to (1) generate an Analysis Image corresponding to the source image, (2) generate binary images from the Analysis Image, (3) model the binary images with fractal mathematics, and (4) generate an output signal related to the Original Image. The Analysis Image contains transformed pixel data in which pixels correspond spatially to the Original Image but in which gray-scale planes are more easily segmented and modeled. Binary Images conforming to standards specified in the catalog are constructed as segments of gray-scale planes within the Analysis Image. Binary models are formed with Hutchinson operators suitably constrained to force finite approximations to attractors to conform to standards specified in the catalog. Reversible output signals include Original Image representations having lower prediction-error entropy; irreversible output signals include representations of the Original Image suitable for feature analysis.

Claims

1. A method of analyzing a matrix of pixel values representing an Analysis Image that is a function of pixel values representing an Original Image derived from an image source wherein the matrices representing the Analysis Image and the Original Image are either the same or different, comprising:

(1) selecting (a) a process applicable to a class of images of which the Original Image is member, and (b) specification information for the class of images of which the Original image is a member;

(2) deriving information from the pixel values of the Original Image, and generating a matrix of pixel values for a corresponding Analysis Image as indicated by the specification information selected in step (1);

(3) generating a set of binary images from data representing the Analysis Image data as indicated by the specification information selected in step (1);

(4) modeling the binary images of the set by using fractal geometry using specification information selected in step (1), each binary model being uniquely specified by (a) a set of independent affine transformation parameters, (b) relationships whereby the remaining parameters necessary to uniquely specify affine transformations are derived from the parameters specified in (a), (c) specifications obtained in step (1) and (d) other data derived from the pixel values of the Analysis Image; and

(5) generating a signal containing information about the image source in response to the affine transformation parameters, specifications, and derived data obtained in step (4).

2. The method of claim 1 further comprising storing the data representation of the Original Image obtained in step (5) in memory.

3. The method of claim 1 further comprising transmitting the data representation of the Original Image obtained in step (5) to a local or remote site.

4. The method of claim 1 wherein the data representation is reversible and further comprising reconstructing the Original Image from the generated signal.

5. The method of claim 1 wherein the process and specific information are selected from a data library where they are stored.

6. The method of claim 5 further including updating the specification information in the data library.

7. The method of claim 5 wherein the data library is operated so the matrix of pixel values representing the Analysis Image is obtained from the matrix of pixel values representing the Original Image by reading a coded procedure stored in the data library.

8. The method of claim 5 wherein the data library is operated so at least one "standard" fractal-like form of the Analysis Image is read from a catalog.

9. The method of claim 8 wherein the data library is operated so the Analysis Image is segmented by a coded method according to locations of the fractal-like forms.

10. The method of claim 9 wherein the data library is operated so a set of binary images is obtained for each segment by the coded method.

11. The method of claim 5 wherein the data library is operated so models of the binary images are derived by a coded method.

12. The method of claim 5 wherein the data library is operated so a specification of output data is derived.

13. The method of claim 1 wherein the Original Image is a gray-scale level pixel representation of a predetermined type of image source and the Analysis Image is a difference image in which pixels of the Analysis Image spatially correspond to pixels in the Original Image, and further comprising assigning numeric values to each pixel in the interior of the Analysis Image as the gray-scale difference between the values of (a) said each pixel in the Original Image and (b) the weighted average gray-scale values of N pixels of the Original Image immediately adjacent, in directions 1 through N, said each pixel in the Original Image, where the directions 1 through N are not co-linear.

14. The method of claim 7 further comprising identifying subimages corresponding to entries of the catalog.

15. A method of processing data derived from an image source included in a predetermined class of images comprising

deriving a matrix of pixel values representing an analysis image that is a function of pixel values representing an original image derived from the image source,

generating a set of binary images from data representing the analysis image,

modeling the binary images of the set by using fractal geometry, each binary model being uniquely specified by a set of independent affine transformation parameters and relationships whereby the remaining parameters necessary to uniquely specify affine transformations that may be derived, information for the class of images and data derived from the pixel values of the analysis image;

and

deriving a diagnostic indicator of the set of binary images by generating signals containing information about coordinates of the source image and the deviation of the source image from the modeled binary images uniquely specified by the affine transformation parameters, the information for the class of images and the data derived from the pixel values of the analysis image, the diagnostic indicator being a fractal dimensionality parameter of the set of binary images.

22. The method of claim 15 wherein pixel values in the Analysis Image are derived from the threshold cutoff values of corresponding pixels in the Original Image, and the set of binary images is derived from the Original Image by:

generating at least one threshold cutoff binary image for each of multiple gray-scale levels of the Analysis Image, and modeling the at least one threshold cutoff binary image for each-gray-scale level by using fractal mathematics.

23. The method of claim 22 further including deriving information about the Original Image by generating signals representing affine transformations of each modeled cutoff binary image, a measure of similarity between the model and the Analysis Image, and spatial coordinates of the binary image.

24. The method of claim 15 wherein the Analysis Image differs from the Original Image and further including changing the amplitudes of features of the Original Image as a function of the Analysis Image, and generating a signal indicating the procedure for changing the amplitudes of features.

38. The method of claim 15 wherein the diagnostic indicator provides a measure of lacunarity of the set of binary images.

39. The method of claim 15 wherein the diagnostic indicator is a texture analysis of the set of binary images.

16. A method of processing data derived from an image source included in a predetermined class of images comprising

deriving a matrix of pixel values representing an analysis image that is a function of pixel values representing an original image derived from the image source,

generating a set of binary images from data representing the analysis image

modeling the binary images of the set by using fractal geometry, each binary model being uniquely specified by a set of independent affine transformation parameters and relationships whereby the remaining parameters necessary to uniquely specify affine transformations that may be derived, information for the class of images and data derived from the pixel values of the analysis image;

and

generating a signal containing information about the source image from the modeled binary images uniquely specified by the affine transformation parameters, the information for the class of images and the data derived from the pixel values of the analysis image,

the original image being a gray-scale level pixel representation of a predetermined type of image source and the analysis image being a difference image in which pixels of the analysis image spatially correspond to pixels in the original image, and further comprising assigning numeric values to each pixel in the interior of the analysis image as the gray-scale difference between the values of (a) said each pixel in the original image and (b) the weighted average gray-scale values of N pixels of the original image immediately adjacent, in directions 1 through N, said each pixel in the original image, where the directions 1 through N are not co-linear, where N is an integer greater than 1.

17. The method of claim 16 further comprising locating a boundary between different assigned numeric values in the Analysis Image, the boundary as a function of spatial locations of assigned numeric values having a common value.

18. The method of claim 17 further comprising defining the boundary by a small number of affine transformation parameters.

19. The method of claim 18 further comprising combining signals representing the affine transformation parameters and an indication of the boundary location to transform pixel values of the Original Image into the binary image models.

20. The method of claim 19 further comprising modifying the Original Image with each binary image model to reduce entropy of a prediction error distribution derived from the Original Image.

21. The method of claim 20 wherein entropy is reduced by changing in a predetermined direction all pixels of the Original Image on one side of the boundary by an amount equal to the common value from which the boundary of the Analysis Image is derived.

25. A computer memory for use with a computer analyzing a matrix of pixel values representing an analysis image that is a function of pixel values representing an original image derived from an image source, the matrices representing the analysis image and the original image being either the same or different, the computer memory storing signals causing the computer to:

(1) select (a) a process applicable to a class of images of which the original image is a member, and (b) specification information for the class of images of which the original image is a member;

(2) derive information from the pixel values of the original image, and generate a matrix of pixel values for a corresponding analysis image as indicated by the specification information selected in operation (1);

(3) generate a set of binary images from data representing the analysis image data as indicated by the specification information selected in operation (1);

(4) model the binary images of the set by using fractal geometry using specification information selected in operation (1), each binary model being uniquely specified by (a) a set of independent affine transformation parameters, (b) relationships whereby the remaining parameters necessary to uniquely specify affine transformations are derived from the parameters specified in (a), (c) specifications obtained in operation (1) and (d) other data derived from the pixel values of the analysis image; and

(5) generate a signal containing information about the image source in response to the affine transformation parameters, specifications, and derived data obtained in operation (4).

26. The memory of claim 25 wherein the data representation is reversible and the memory stores signals to cause the computer to reconstruct the original image from the generated signal.

27. The memory of claim 25 wherein the memory stores signals to cause the computer to store in the memory further signals indicative of:

(a) the original image as a gray-scale level pixel representation of a predetermined type of image source and (b) the analysis image as a difference image in which pixels of the analysis image spatially correspond to pixels in the original image; and

the memory stores signals causing the computer to assign numeric values to each pixel in the interior of the analysis image as the gray-scale difference between the values of (a) said each pixel in the original image and (b) the weighted average gray-scale values of N pixels of the original image immediately adjacent, in directions 1 through N, of said each pixel in the original image, where the directions 1 through N are not co-linear.

28. A computer memory for use with a computer that processes data derived from an image source included in a predetermined class of images, the computer memory storing signals for causing the computer to:

(1) derive and store a matrix of pixel values representing an analysis image that is a function of pixel values representing an original image derived from the image source,

(2) generate a set of binary images from data representing the analysis image,

(3) model the binary images of the set by using fractal geometry, each binary model being uniquely specified by a set of independent affine transformation parameters and relationships whereby the remaining parameters necessary to uniquely specify affine transformations may be derived, information for the class of images and data derived from the pixel values of the analysis image; and

(4) derive a diagnostic indicator of the set of binary images by generating signals containing information about coordinates of the source image and the deviation of the source image from the modeled binary images uniquely specified by the affine transformation parameters, the information for the class of images and the data derived from the pixel values of the analysis images, the diagnostic indicator being a fractal dimensionality parameter of the set of binary images.

35. The memory of claim 28 wherein the memory stores other signals causing the computer to derive pixel values in the analysis image from the threshold cutoff values of corresponding pixels in the original image, and derive the set of binary images from the original image by:

generating at least one threshold cutoff binary image for each of multiple gray-scale levels of the analysis image, and model the at least one threshold cutoff binary image for each gray-scale level by using fractal mathematics.

36. The memory of claim 35 wherein the memory stores additional signals causing the computer to derive information about the original image by generating signals representing affine transformations of each modeled cutoff binary image, a measure of similarity between the model and the analysis image, and spatial coordinates of the binary image.

37. The memory of claim 28 wherein the analysis image differs from the original image and the memory stores added signals causing the computer to change the amplitudes of features of the original image as a function of the analysis image, and generate a signal indicating the procedure for changing the amplitudes of features. the interior of the analysis image as the gray-scale difference between the values of (a) said each pixel in the original image and (b) the weighted average gray-scale values of N pixels of the original immediately adjacent image, in directions 1 through N, of said each pixel in the original image, where the directions 1 through N are not co-linear, where N is an integer greater than 1.

40. The memory of claim 28 wherein the diagnostic indicator provides a measure of lacunarity of the set of binary images.

41. The memory of claim 28 wherein the diagnostic indicator is a texture analysis of the set of binary images.

29. A computer memory for use with a computer that processes data derived from an image source included in a predetermined class of images, the computer memory storing signals for causing the computer to:

(1) derive and store a matrix of pixel values representing an analysis image that is a function of pixel values representing an original image derived from the image source,

(2) generate a set of binary images from data representing the analysis image,

(3) model the binary images of the set by using fractal geometry, each binary model being uniquely specified by a set of independent affine transformation parameters and relationships whereby the remaining parameters necessary to uniquely specify affine transformations may be derived, information for the class of images and data derived from the pixel values of the analysis image; and

(4) generate a signal containing information about the source image from the modeled binary images uniquely specified by the affine transformation parameters, the information for the class of images and the data derived from the pixel values of the analysis image,

the original image being a gray-scale level pixel representation of a predetermined type of image source and the analysis image being a difference image in which pixels of the analysis image spatially correspond to pixels in the original image, and the memory storing further signals causing the computer to assign and store in the memory numeric values to each pixel in the interior of the analysis image as the gray-scale difference between the values of (a) said each pixel in the original image and (b) the weighted average gray-scale value of N pixels of the original immediately adjacent image, in direction 1 though N, of said each pixel in the original image, where the direction 1 through N are co-linear, where N is an integer greater than 1.

30. The memory of claim 29 wherein the memory stores additional signals causing the computer to locate a boundary between different assigned numeric values in the analysis image, the boundary as a function of spatial locations of assigned numeric values having a common value.

31. The memory of claim 30 wherein the memory stores other signals causing the computer to define the boundary by a small number of affine transformation parameters.

32. The memory of claim 31 wherein the memory stores additional signals causing the computer to combine signals representing the affine transformation parameters and an indication of the boundary location to transform pixel values of the original image into the binary image models.

33. The memory of claim 32 wherein the memory stores added signals causing the computer to modify the original image with each binary image model to reduce entropy of a prediction error distribution derived from the original image.

34. The memory of claim 33 wherein the memory stores still further signals causing the computer to reduce entropy by changing in a predetermined direction all pixels of the original image on one side of the boundary by an amount equal to the common value from which the boundary of the analysis image is derived.