Fractal model based image processing

• US 4,789,933 A
• Filed: 02/27/1987
• Issued: 12/06/1988
• Est. Priority Date: 02/27/1987
• Status: Expired due to Fees

• Current Assignee: Picker International, Inc. (Marconi Holdings Company Limited)
• Original Assignee: Picker International, Inc. (Marconi Holdings Company Limited)
• Inventors: Tesic, Mike M., Chen, Victor C.
• Primary Examiner(s): NOT, DEFINED
• Assistant Examiner(s): Kibby, Steven G.
• Application Number:
  US07/019,568
• Time in Patent Office: 648 Days
• Field of Search: 382/27, 382/47, 382/54, 358/166, 358/167, 364/414
• US Class Current: 382/128
• CPC Class Codes: G06T 5/20 using local operators

Abstract

A medical diagnostic apparatus (A) generates medical diagnostic data which is reconstructed by an imager (B) into an electronic image representation. The electronic image representation includes an array of digital pixel values which represent a gray scale intensity of a man-readable image displayed on a video monitor (62). An image improving circuit (C) replaces each pixel value I(i,j) with an improved pixel value I'"'"'(i,j) defined as follows:

I'"'"'(i,j)=G(i,j)[I(i,j)-I(i,j)]+I(i,j),

where G(i,j) is a transfer function uniquely defined for each pixel and I is the mean of pixel values of neighboring pixels. The transfer function is based on a self-similarity value which is derived by comparing (i) a variation between the pixel value I(i,j) and pixel values in a first surrounding ring with (ii) a variation between the pixel value I(i,j) and pixel values in a second, larger surrounding ring. A zoom circuit (D) enlarges a selected portion of an improved image. Empty or unfilled pixel values I(inter) are interpolated from a combination of an average of neighboring pixel values I(avg) and a fractal value. The fractal value is a combination of a random number and a weighting factor R determined in accordance with the average self-similarity value of the neighboring filled pixels, i.e.:

I(inter)=I(avg)+R·(random#).

Claims

1. An apparatus for generating medical diagnostic image representations, the apparatus comprising:

a medical diagnostic apparatus for generating diagnostic data indicative of a preselected image region of a patient;

an imager for generating an electronic image representation from the diagnostic data, the electronic image representation including a pixel value corresponding to each pixel of a pixel array, each pixel value is indicative of an image property of a corresponding subregion of the image region of the patient;

a self-similarity value generating means for generating a self-similarity value corresponding to each pixel, the self-similarity value generating means being operatively connected with the imager to receive pixel values therefrom, each self-similarity value varying in accordance with a ratio of (i) a difference between the corresponding pixel value and a first set of pixel values contiguous to and surrounding the corresponding pixel value and (ii) the corresponding pixel value and a second set of pixel values contiguous to and surrounding the first set of pixel values;

an image improvement means for replacing each pixel value by a combination of the replaced pixel value and an average of surrounding pixel values, the combination being weighted in accordance with the corresponding self-similarity value, the image improvement means being operatively connected with the imager and the self-similarity value generating means.

2. An apparatus for generating medical diagnostic image representations, the apparatus comprising:

a medical diagnostic apparatus for generating medical diagnostic signals indicative of an image region of a patient;

an imager for transforming the medical diagnostic signals into electronic image representations, each electronic image representation including a pixel value corresponding to each pixel of a pixel array;

a loading means for loading each pixel value into uniformly distributed, spaced pixels of a zoom memory means such that the pixel values are stored in uniformly distributed filled pixels and empty pixels are disposed therebetween;

an interpolating means for generating a pixel value for each empty pixel, the interpolating means including;

an averaging means for generating an average of neighboring pixel values;

a random number generator for generating a random number;

a self-similarity value generating means for generating a self-similarity value corresponding to each pixel, the self-similarity value generating means being operatively connected with the imager to receive pixel values therefrom, each self-similarity value varying in accordance with a ratio of differences between the corresponding pixel value and pixel values in at least two rings contiguous to the corresponding pixel value;

a weighting means for weighting the random number in accordance with the self-similarity value, the weighting means being operatively connected with the self-similarity value generating means and the random number generator; and

,

combining means for combining the pixel value average from the averaging means and the weighted randon number from the weighting means, the combining means being operatively connected with the averaging means to receive the average value therefrom, with the weighting means for receiving the weighted random number therefrom, and with the zoom memory means for storing the combined average pixel value and weighted random number at the corresponding empty pixel thereof.

3. A method of medical diagnostic imaging comprising:

converting medical diagnostic data into an electronic image representation which includes an array of pixel values;

generating a self-similarity value corresponding to each pixel value of the electronic image representation, each self-similarity value being a dimensionless value which varies with a degree of irregularity among pixel values in a neighborhood of the corresponding pixel value;

providing an improved electronic image representation by replacing each pixel value by a combination of the replaced pixel value and an average of surrounding pixel values, the combination being weighted in accordance with the corresponding self-similarity value.

4. The method as set forth in claim 3 wherein the self-similarity value generating step includes:

determining a difference between the corresponding pixel value and each pixel value within a nearest neighbor first ring and averaging the first ring differences;

determining a difference between the corresponding pixel value and pixel values in a next nearest neighboring second ring and averaging the second ring differences.

5. The method as set forth in claim 4 wherein the self-similarity value generating step further includes determining a logarithm of a ratio of the first ring and second ring differences.

6. The method as set forth in claim 5 further including multiplying the logarithm by a constant selected in accordance with the diameter of the rings.

7. The method as set forth in claim 5 wherein the replacing step further includes determining a transfer function for each pixel value from the corresponding self-similarity value and wherein the combination is weighted by multiplying the combination by the transfer function.

8. The method as set forth in claim 7 wherein the transfer function is inversely proportional to the self-similarity value for at least a selected range of self-similarity values.

9. The method as set forth in claim 7 wherein the transfer function is set equal to a first constant when the self-similarity value exceeds a preselected value and the transfer function is set equal to a second preselected constant when the self-similarity value is equal to or less than the preselected value.

10. The method as set forth in claim 4 further including enlarging a selected portion of the electronic image representation, the enlarging step including distributing pixel values from the selected image representation portion substantially uniformly among available pixels of an enlarged image electronic representation such that a fraction of the enlarged image representation pixels are filled by the transferred pixel values and empty pixels are defined therebetween and interpolating neighboring filled pixel values to derive pixel values for the empty pixels.

11. The method as set forth in claim 10 wherein the interpolating step includes:

for each empty pixel, averaging pixel values from neighboring filled pixels; and

,

adding to the averaged pixel valves a fractal value which varies in accordance with a random number and the self-similarity value.

12. The method as set forth in claim 11 wherein the fractal value varies in proportion to a constant raised to a power of an average of the self-similarity values corresponding to the neighboring filled pixels.

13. A method of improving an image comprising:

converting data into an electronic image representation which includes an array of pixel values;

determining a first difference between the corresponding pixel value and each pixel value within a first surrounding ring and averaging the first ring differences;

determining a difference between the corresponding pixel value and each pixel value in a second surrounding ring and averaging the second ring differences;

generating a self-similarity value corresponding to each pixel value of the electronic image representation in accordance with a difference between the first and second difference averages corresponding to the same pixel value;

providing an improved electronic image representation by replacing each pixel value by a combination of the replaced pixel value, pixel values of pixels surrounding the replaced pixel value, and the self-similarity value corresponding to the replaced pixel value.

14. A method of medical diagnostic imaging comprising:

converting medical diagnostic data into an electronic image representation which includes an array of pixels, each pixel having a pixel value;

generating a self-similarity value corresponding to each pixel;

enlarging a selected portion of the electronic image representation, the enlarging step including distributing pixel values of the selected image representation portion substantially uniformly among pixels of an enlarged image electronic representation such that a fraction of the enlarged image representation pixels are filled by the transferred pixel values and empty pixels are defined therebetween;

for each empty pixel, averaging pixel values from neighboring filled pixels;

for each empty pixel deriving a fractal value which varies in accordance with a random number and the self-similarity values of the neighboring filled pixels; and

,

combining the fractal value and the neighboring pixel value average that corresponds to the same empty pixel.

15. The method as set forth in claim 14 wherein the fractal value varies in proportion to a constant raised to a power of an average of the self-similarity values corresponding to the neighboring filled pixels.

16. The method as set forth in claim 15 wherein each self-similarity value varies in accordance with a ratio of variations between the corresponding pixel value and pixel values in at least two surrounding rings.

17. The method as set forth in claim 15 wherein the self-similarity value generating step includes:

determining a difference between the corresponding pixel value and each pixel value within a first surrounding ring and averaging the first ring differences;

determining a difference between the corresponding pixel value and each pixel value in a second surrounding ring and averaging the second ring differences.

18. The method as set forth in claim 17 wherein the self-similarity value generating step further includes determining a logarithm corresponding to a ratio of the first ring and second ring differences.

19. The method of claim 18 including providing an improved electronic image representation by replacing each pixel value by a combination of the pixel value and an average of surrounding pixel values, each combination being weighted in accordance with the corresponding self-similarity value.

20. A method of enlarging a selected portion of an image representation comprising:

converting data into an electronic image representation which includes an array of pixels, each pixel having a pixel value;

generating a self-similarity value corresponding to each pixel from pixel values of adjacent pixels;

distributing pixel values of the selected image representation portion substantially uniformly among pixels of an enlarged image electronic representation such that a fraction of the enlarged image representation pixels are filled by the transferred pixel values and empty pixels are defined therebetween;

for each empty pixel, averaging pixel values from neighboring filled pixels;

for each empty pixel, deriving a fractal value which varies in accordance with a random number and the self-similarity values of adjacent pixels; and

,

combining the fractal value and the neighboring pixel value average that corresponds to the same empty pixel.