Methods and apparatus for image compression by iterated function system

• US 4,941,193 A
• Filed: 10/02/1987
• Issued: 07/10/1990
• Est. Priority Date: 10/02/1987
• Status: Expired due to Term

• Current Assignee: MediaBin, Inc. (HP Inc.)
• Original Assignee: MediaBin, Inc. (HP Inc.)
• Inventors: Sloan, Alan D., Barnsley, Michael F.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Couso, Jose L.
• Application Number:
  US07/104,412
• Time in Patent Office: 1,012 Days
• Field of Search: 382/1, 382/41, 382/44, 382/56, 340/750, 340/798, 340/799, 364/514, 364/518, 364/522, 364/523
• US Class Current: 382/249
• CPC Class Codes: G06T 9/001 Model-based coding, e.g. wire frame
• CPC Class Codes: H04N 19/99 involving fractal coding

Abstract

A method and apparatus for obtaining highly compressed images employing an iterated function system (IFS). An original input or target image is subdivided into regions having similar characteristics. Contractive copies or maps of a particular region, which are the results of affine transformations to the region, are generated and tiled with respect to the input image until the entire region is covered and a collage is formed. Each region is processed in like manner. The affine transformation coefficients or IFS codes completely represent the input image, and are stored or transmitted. To generate an image from the IFS codes, a decoding system is disclosed. One disclosed method involves a chaotic dynamical system. A random iteration of the IFS codes is performed until an attractor, which is the target image, emerges and stabilizes. Another disclosed deterministic method repeatedly and successively applies the IFS codes to an arbitrary starting image until the attractor emerges. Also disclosed are various methods for representing and compressing the color information of an image, including a method for employing an additional spatial dimension in the mappings and a method for employing an arbitrary probabilistic measure for the color rendering.

Claims

1. A method for compressing images, comprising the steps of:

providing an input image having at least two dimensions;

storing the image in a memory;

displaying the image stored in the memory on a display;

selecting a geometric first region of the image displayed on the display with selecting means;

generating and displaying a contractive copy of the first region in the memory;

storing a set of transformation coefficients corresponding to the transformation creating the contractive copy;

on the display, collaging the transformed contractive copy with respect to the first region to align geometric features of the transformed contractive copy with corresponding geometric features of the first region;

repeating the steps of generating and displaying a contractive copy, storing the transformation coefficients, and collaging the transformed contractive copy until substantially all of the first region is covered by a collage of transformed contractive copies, thereby providing a set of a plurality of transformation coefficients; and

providing as an output a set of iterated function system compression codes corresponding to the set of a plurality of transformation coefficients.

2. The method of claim 1, wherein the transformation coefficients are affine transformation coefficients.

3. The method of claim 1, wherein the step of collaging the transformed copy on the display includes the step of modifying the transformation coefficients corresponding to the coordinates of the location and dimensions of the transformed copy.

62. The method of claim 1, wherein the method is performed on a programmed computer.

63. The method of claim 1, further comprising the step of transmitting or storing the transformation coefficients as data compression codes representative of the input image.

64. The method of claim 1, further comprising the steps of:

comparing the input image to an approximation image stored in an approximation image memory;

providing error data as a function of the error between the input image and the approximation image; and

modifying the collage of contractive transformed copies in response to the error data to minimize the number of transformation coefficients.

65. The method of claim 63, wherein the step of comparing the approximation image to the input image is performed by an operator viewing the two images on a display screen, and wherein the step of modifying the collage comprises the steps of generating and collaging additional transformed copies or repositioning or reshaping previously-generated transformed copies with respect to the collage.

66. The method of claim 64, wherein the method is carried out separately for each of a plurality of portions of the input image, thereby generating a plurality of sets of transformation coefficients, each of said sets of transformation coefficients corresponding to the carrying out of the method for a single portion of the input image, and further comprising the step of providing residual image data representing those portions of the input image not represented by the transformation coefficients.

67. The method of claim 66, wherein said residual image data comprises data corresponding to the carrying out of the method upon portions of the image remaining after carrying out of the method for other portions of the image.

68. The method of claim 66, wherein said residual image data comprises data corresponding to the location and color or gray scale rendering of portions of the image remaining after carrying out the method for other portions of the image.

69. The method of claim 1, wherein the step of collaging the transformed copy on the display includes the step of modifying the transformation coefficients corresponding to the coordinates of the location and dimensions of the transformed copy.

4. A random iteration method for decoding iterated function system compression codes to generate an image, comprising the steps of:

receiving and storing a plurality of compression codes in a compression code memory;

determining an initial point in a display memory;

randomly selecting one of the compression codes in the compression code memory;

performing an affine transformation on the initial point corresponding to the affine transformation represented by the selected compression code to move the initial point to a second point in the display memory;

storing a picture element in the display memory at a memory location corresponding to the location of the second point;

utilizing the second point from the preceding step as an initial point, repeating the steps of randomly selecting one of the compression codes, performing an affine transformation, and storing a picture element, until an output image converges to an attractor in the display memory; and

displaying data corresponding to the stored picture elements in the display memory corresponding to the attractor as a decoded output image.

5. The method of claim 4, wherein the method includes the step of discarding the stored picture elements generated by a first predetermined number n of iterations so as to prevent display of points substantially distant from the attractor.

6. The method of claim 5, wherein n is at least 50.

7. A deterministic method for decoding iterated function system compression codes to generate an image on a graphics display terminal, comprising the steps of:

receiving and storing a plurality of n sets of compression codes;

providing an initial image in a display memory associated with the graphics display terminal;

performing transformations on the initial image in the memory corresponding to each of the n sets of compression codes;

combining the results of the transformations so as to obtain a second image in the memory;

utilizing the second image as the initial image, repeatedly iterating by carrying out the above steps, for a plurality of iterations of the n sets of compression codes, until an attractor stabilizes;

storing in the memory a plurality of picture elements corresponding to the attractor as a final image; and

displaying on the graphics display terminal data corresponding to the stored picture elements of the final image corresponding to the attractor as a decoded output image.

8. The method of claim 7, wherein the step of combining the results of the transformations comprises taking a weighted average of the outcome of the individual transformations with respect to each of the plurality of picture elements.

9. The method of claim 7, wherein the step of storing picture elements comprises:

storing a set of picture elements corresponding to the results of the transformation step in a buffer memory to obtain the second image; and

replacing the contents of the display memory with the contents of the buffer memory after completion of the transformation step.

10. Apparatus for decoding iterated function system compression codes to generate an image by random iteration, comprising:

a compression code memory for storing a plurality of compression codes;

a display memory for storing an image being generated;

means for determining an initial point in said display memory;

means for randomly selecting one of said compression codes in said compression code memory;

means for performing an affine transformation on said initial point corresponding to the affine transformation represented by said selected compression codes to move said initial point to a second point;

means for storing a picture element in a display memory at a memory location in said display memory corresponding to the location of said second point;

iteration means for repeatedly utilizing said second point as said initial point for subsequent operations of said compression code selecting means, said affine transformation means, and said picture element storing means, until an output image converges to an attractor in said display memory; and

means for displaying data stored in said display memory corresponding to stored picture elements corresponding to said attractor as a decoded output image.

11. The apparatus of claim 10, further comprising means for discarding stored picture elements generated by a first predetermined number n of iterations so as to prevent display of points substantially distant from said attractor.

12. The apparatus of claim 11, wherein n is at least 50.

13. An apparatus for decoding iterated function system compression codes to generate an image by deterministic decoding, comprising:

a compression code memory for storing a plurality of n sets of compression codes;

a initial image display memory for storing an initial image;

a generated image display memory for storing a generated image;

means for providing an initial image in said initial image display memory;

means for performing an affine transformation on said initial image in said initial image display memory corresponding to each of said sets of compression codes;

means for combining the results of said affine transformations to produce a second image in said generated image display memory;

means for transferring said second image from generated image display memory to initial image display memory;

deterministic iteration means for causing said affine transformation means, said combining means, and said transferring means to repeatedly operate for a plurality of iterations of the n set of compression codes, until an attractor stabilizes;

means for storing a plurality of picture elements corresponding to the attractor in said display memory; and

means for displaying data corresponding to the stored picture elements stored in said display memory corresponding to said attractor as a decoded output image.

14. The apparatus of claim 13, wherein said combining means comprises means for taking a weighted average of the results of the transformations with respect to each of the plurality of picture elements, with averaging occurring independently at each pixel location.

15. The apparatus of claim 13, wherein said picture element storing means comprises:

a buffer memory for storing an intermediate image;

means for storing a set of picture elements corresponding to the results of said affine transformation means in said buffer memory to obtain the second image; and

means for replacing the contents of said display memory with the contents of said buffer memory.

16. An apparatus for decoding a compressed image represented by a plurality of sets of compression codes, each one of said sets of compression codes having a contractivity factor S, where 0≦

S<

1, and each one of said sets of compression codes having a probability code associated therewith, comprising;

a compression code memory for storing a plurality of sets of input compression codes;

a probability memory for storing the probability codes associated with each one of said sets of compression codes;

means for storing said input compression codes in said compression code memory in addressable locations having addresses corresponding to said probability codes;

a display memory corresponding to a display space for storing an image;

a random number generator for generating a probability addresses which randomly vary;

compression code selecting means responsive to said probability addresses for selecting compression codes stored at addresses in said compression code memory at a frequency corresponding to said probability codes, such that a compression code having a higher probability code is more likely to be selected than a compression code having a lower probability code;

matrix processing means responsive to compression codes provided by said compression code selecting means for performing an approximately contractive affine transformation on an initial point in said display space to determine a second point in said display space;

means for storing image data in said display memory corresponding to said second point in said display space;

control means for causing said compression code selecting means and said matrix processing means to iteratively plot a plurality of points in said display space until an attractor of said compression codes stabilizes;

means for storing data in said display memory corresponding to points plotted in said display space by said control means; and

means for displaying the contents of said display memory after a plurality of iterations.

17. The apparatus of claim 16, further comprising hit count means for providing color rendering of images stored in said display memory, comprising:

a hit count memory for storing a hit count of the number of times that each point in said display space was plotted;

hit counter means responsive to the plotting of a particular point in display space for incrementing the hit count corresponding to that particular point in said display memory;

means for retrieving the hit count for each point in said display memory and providing a color index;

a color index memory for storing color values corresponding to a plurality of said color indices;

means for displaying the color value stored in said color index memory associated with the color index for a particular point when said particular point is displayed by said displaying means.

18. The apparatus of claim 16, further comprising means for assigning colors to a plurality of hit count intervals, comprising:

means for selecting a plurality n+1 of hit count values so as to define n hit count intervals; and

means for associating each of n colors with a particular one of said hit count intervals.

19. The apparatus of claim 16, wherein each of said sets of compression codes corresponds to a three-dimensional contractive affine transformation of portion of an original input image, two of the dimensions corresponding to spatial dimensions and a third dimension corresponding to a color dimension, wherein said display space is as three-dimensional display space, and further comprising added dimension means for providing color rendering of images stored in said display memory, comprising:

a pixel memory for storing color data corresponding to a color value for each point in a display space;

means responsive for storing said color data in said pixel memory in response to a three-dimensional affine transformation performed by said matrix processing means;

means for retrieving said color data from said pixel memory for each point in said display memory; and

means for displaying said color data for a particular point when said particular point is displayed by said displaying means.

20. Apparatus for compressing an image represented by an array of pixels, comprising:

a target memory means for storing an uncompressed input target image;

means for generating, aligning, and overlaying a collage comprising a plurality of contractive affinely transformed copies of a portion of said target image with respect to said portion of said target image stored in said target memory means until particular geometric features of said portion of said target image are substantially covered by said collage;

output means for providing, as a compressed image output, a plurality of compression codes corresponding to the numerical coefficients of said plurality of contractive affinely transformed copies of said portion of said target image;

decoder means responsive to said compression codes for generating and storing an approximation image in said approximation memory means;

an approximation image memory means for storing said approximation image;

comparator means for comparing said target image stored in said target memory means to said approximation image stored in said approximation image memory means and for providing error data;

affine transformation correction means responsive to said error data for modifying said collage of contractive affinely transformed copies of said portion of said target image to minimize the number of said compression codes.

21. The apparatus of claim 20, wherein said decoder means comprise a deterministic decoder.

22. The apparatus of claim 21, wherein said deterministic decoder comprises:

a compression code memory for storing a plurality of n sets of compression codes;

a initial image display memory for storing an initial image;

a generated image display memory for storing a generated image;

means for providing an initial image in said initial image display memory;

means for performing an affine transformation on said initial image in said initial image display memory corresponding to each of said sets of compression codes;

means for combining the results of said affine transformations to produce a second image in said generated image display memory;

means for transferring said second image from generated image display memory to initial image display memory;

deterministic iteration means for causing said affine transformation means, said combining means, and said transferring means to repeatedly operate for a plurality of iterations of the n sets of compression codes, until an attractor stabilizes;

means for storing a plurality of picture elements corresponding to the attractor in said display memory; and

means for displaying data corresponding to the stored picture elements stored in said display memory corresponding to said attractor as a decoded output image.

23. The apparatus of claim 22, wherein said combining means comprises means for taking a weighted average of the results of the transformations with respect to each of the plurality of picture elements, with averaging occurring independently at each pixel location.

24. The apparatus of claim 22, wherein said picture element storing means comprises:

a buffer memory for storing an intermediate image;

means for storing a set of picture elements corresponding to the results of said affine transformation means in said buffer memory to obtain the second image; and

means for replacing the contents of said display memory with the contents of said buffer memory.

25. The apparatus of claim 20, wherein said decoder comprises a random iteration decoder.

26. The apparatus of claim 25, wherein said random iteration decoder comprises:

a compression code memory for storing a plurality of compression codes;

a display memory for storing an image being generated;

means for determining an initial point in said display memory;

means for randomly selecting one of said compression codes in said compression code memory;

means for performing an affine transformation on said initial point corresponding to the affine transformation represented by said selected compression codes to move said initial point to a second point;

means for storing a picture element in a display memory at a memory location in said display memory corresponding to the location of said second point;

iteration means for repeatedly utilizing said second point as said initial point for subsequent operations of said compression code selecting means, said affine transformation means, and said picture element storing means, until an output image converges to an attractor in said display memory; and

means for displaying data stored in said display memory corresponding to stored picture elements corresponding to said attractor as a decoded output image.

27. The apparatus of claim 26, wherein further comprising means for discarding stored picture elements generated by a first predetermined number n of iterations so as to prevent display of points substantially distant from said attractor.

28. The apparatus of claim 27, wherein n is at least 50.

29. A method of compressing an image represented by an array of pixels, comprising:

storing an uncompressed input target image in a target memory;

generating contractive transformed copies of a portion of the target image;

aligning the contractive transformed copies with respect to the portions of said target image stored in the target memory until particular geometric features of the portions of the target image are substantially covered by a collage;

providing, as a compressed image output, a plurality of compression codes corresponding to the numerical coefficients of the plurality of contractive transformed copies of the portion of the target image;

decoding the compression codes to generate approximation image;

storing the approximation image in an approximation image memory;

comparing the target image stored in the target memory to the approximation image stored in the approximation image memory means and providing error data as a function of the error between the two images;

modifying the collage of contractive transformed copies of the portion of the target image in response to the error data to minimize the number of compression codes.

30. The method of claim 29, wherein the method is carried out on a programmed computer.

31. The method of claim 29, wherein the step of comparing the approximation image to the target image is performed by an operator viewing the two images on a display screen, and wherein the step of modifying the collage comprises the step generating and collaging additional transformed copies or the step of repositioning or reshaping previously-generated transformed copies with respect to the collage.

32. The method of claim 29, wherein the method is carried out separately for each of a plurality of portions of the target image, thereby generating a plurality of sets of compression codes, each of said sets of compression codes corresponding to the carrying out of the method for a single portion of the target image, and further comprising the step of providing residual image data representing those portions of the target image not represented by compression codes.

33. The method of claim 32, wherein said residual image data comprises data corresponding to the carrying out of the method upon portions of the image remaining after carrying out of the method for other portions of the image.

34. The method of claim 32, wherein said residual image data comprises data corresponding to the location and color or gray scale rendering of portions of the image remaining after carrying out the method for other portions of the image.

35. The method of claim 29, wherein the transformation coefficients are affine transformation coefficients.

36. The method of claim 29, wherein the step of collaging the transformed copy on the display includes the step of modifying the transformation coefficients corresponding to the coordinates of the location and dimensions of the transformed copy.

37. In an image compression system, a method for encoding rendering information for an image represented by an array of pixels, each pixel having a rendering value, comprising the steps of, in a programmed computer:

determining the number of different rendering values in an input image to be encoded;

determining a measure corresponding to the cumulation of the rendering values for all pixels in the image;

assigning a weighting to each one of the number of different rendering values;

assigning a rendering mapping of each one of the rendering values to one of the different renderings to be encoded;

encoding predetermined features of the input image with an iterated function system to as to obtain a plurality of sets of transformation coefficients representing the image without color rendering;

assigning a probability code to each one of the sets of transformation coefficients in relation to the relative weight assigned to the number of different rendering values; and

providing as an encoded output a set of compression codes comprising a plurality of sets of transformation coefficients and a corresponding probability code for each one of the sets of transformation coefficients.

38. The method of claim 37, wherein the step of encoding predetermined features of the input images comprises encoding geometric features of the image having common characteristics.

39. The method of claim 37, wherein the step of assigning a rendering mapping comprises assigning an arbitrary assignment of a predetermined color to each of the rendering values.

40. The method of claim 37, wherein the step of encoding predetermined features of the input images comprises the steps of:

storing the input image in a memory;

displaying the image stored in the memory on a display;

selecting a geometric first region of the image displayed on the display with selecting means;

generating and displaying a contractive copy of the first region in the memory;

storing a set of transformation coefficients corresponding to the transformation creating the contractive copy;

on the display, collaging the transformed contractive copy with respect to the first region to align geometric features of the transformed contractive copy with corresponding geometric features of the first region;

repeating the steps of generating and displaying a contractive copy, storing the transformation coefficients, and collaging the transformed contractive copy until substantially all of the first region is covered by a collage of transformed contractive copies, thereby providing a set of a plurality of transformation coefficients; and

providing as an output a set of iterated function system compression codes corresponding to the set of a plurality of transformation coefficients.

41. The method of claim 40, wherein the transformation coefficients are affine transformation coefficients.

42. The method of claim 40, wherein the step of collaging the transformed copy on the display includes the step of modifying the transformation coefficients corresponding to the coordinates of the location and dimensions of the transformed copy.

43. The method of claim 37, further comprising a method for decoding encoded rendering information, comprising:

receiving and storing a plurality of compression codes comprising the transformation coefficients and probability codes in a memory;

determining an initial point in a display memory;

randomly selecting one of the transformation coefficients in the memory weighted by its corresponding probability code;

performing a transformation on the initial point corresponding to the selected transformation coefficients to move the initial point to a second point;

storing a picture element in a display memory at a memory location corresponding to the location of the second point;

utilizing the second point from the preceding step as an initial point, repeating the steps of randomly selecting one of the transformation coefficients, performing a transformation, and storing a picture element, until an output image converges to an attractor; and

displaying data corresponding to the stored picture elements corresponding to the attractor as a decoded output image.

44. The method of claim 43, further comprising the steps of:

storing in a hit count memory a hit count of the number of times that each point in said display space corresponding to the display memory was plotted;

responsive to the plotting of a particular point in display space, incrementing the hit count corresponding to that particular point in the display space;

after the attractor has stabilized, retrieving the hit count for each point in the display space and providing a rendering index corresponding thereto;

storing in a rendering index memory rendering values corresponding to a plurality of color indices; and

displaying the rendering value stored in the rendering index memory associated with the rendering index for a particular point when that particular point is displayed.

45. The method of claim 37, wherein the rendering information is gray scale information.

46. The method of claim 37, wherein the rendering information is color information.

47. A method for compressing images, comprising the steps of, in a programmed computer:

storing an input target image to be compressed in a target memory;

selecting geometric regions of the input target image having predetermined similar fractal characteristics;

determining a separate set of iterated function system codes corresponding to and representing each of said selected geometric regions, to thereby obtain a plurality of sets of iterated function system codes;

determining a residual set of data corresponding to those residual portions of the image not represented by one of said plurality of sets of iterated function system codes;

encoding the residual set with a conventional data compression scheme to obtain a set of conventional compression codes;

providing as a compressed output representing the entire image a set of iterated function system codes representing the selected geometric regions and a set of conventional compression codes representing the residual portions of the image.

48. The method of claim 47, wherein the iterated function system codes are the coefficients of a collage of contractive transformations applied to each of the selected geometric regions.

49. The method of claim 48, wherein the contractive transformations are affine transformations.

50. The method of claim 47, wherein the iterated function system codes are determined by the steps of:

displaying the image stored in the memory on a display;

generating and displaying a contractive copy of a first geometric region in the memory;

storing a set of transformation coefficients corresponding to the transformation creating the contractive copy;

on the display, collaging the transformed contractive copy with respect to the first region to align geometric features of the transformed contractive copy with corresponding geometric features of the first region; and

repeating the steps of generating and displaying a contractive copy, storing the transformation coefficients, and collaging the transformed contractive copy until substantially all of the first region is covered by a collage of transformed contractive copies, thereby providing a set of a plurality of transformation coefficients.

51. The method of claim 50, wherein the step of collaging the transformed copy on the display includes the step of modifying the transformation coefficients corresponding to the coordinates of the location and dimensions of the transformed copy.

52. The method of claim 47, wherein the method of determining iterated function system codes is an iterative method comprising the steps of:

generating contractive transformed copies of a portion of the target image;

aligning the contractive transformed copies with respect to the selected regions of the target image stored in the target memory until particular geometric features of the selected region of the target image are substantially covered by a collage;

providing, as a compressed image output, a plurality of compression codes corresponding to the numerical coefficients of the plurality of contractive transformed copies of the portion of the target image;

decoding the compression codes to generate approximation image;

storing the approximation image in an approximation image memory;

comparing the target image stored in the target memory to the approximation image stored in the approximation image memory means and providing error data as a function of the error between the two images;

modifying the collage of contractive transformed copies of the portion of the target image in response to the error data to minimize the number of compression codes.

53. The method of claim 52, wherein the method is carried out on a programmed computer.

54. The method of claim 52, wherein the step of comparing the approximation image to the target image is performed by an operator viewing the two images on a display screen, and wherein the step of modifying the collage comprises the step generating and collaging additional transformed copies or the step of repositioning or reshaping previously-generated transformed copies with respect to the collage.

55. The method of claim 52, wherein the method is carried out separately for each of the plurality of regions of the target image, thereby generating a plurality of sets of compression codes, each of said sets of compression codes corresponding to the carrying out of the method for a single region of the target image, and further comprising the step of providing residual image data representing those regions of the target image not represented by compression codes.

56. The method of claim 55, wherein said residual image data comprises data corresponding to the carrying out of the method upon regions of the image remaining after carrying out of the method for other portions of the image.

57. The method of claim 55, wherein said residual image data comprises data corresponding to the location and color or gray scale rendering of portions of the image remaining after carrying out the method for other portions of the image.

58. A deterministic method for decoding iterated function system compression codes to generate an image by parallel processing, comprising the steps of:

receiving and storing a plurality of n sets of compression codes in a decoder;

providing an initial image in a display memory;

concurrently performing in a separate one of a plurality of processing units each of the n transformations corresponding to the stored compression codes on the initial image so as to generate n transformed images,

forming a second image corresponding to the union of the n transformed images;

repeatedly iterating by carrying out the above steps of concurrently performing n transformations and forming the second image, wherein, for each iteration, the second image formed in the previous iteration is utilized as the initial image for the following iteration and continuing such iteration until an attractor image stabilizes; and

displaying data corresponding to the attractor as a decoded output image.

59. The method of claim 58, further comprising the step of partitioning the display memory into m different regions, and wherein the step of performing transformations comprises concurrently performing each of the n transformations on each of the m regions so as to generate n·

m transformed images, and wherein the step of forming a second image comprises forming the second image as the union of n·

m transformed images.

60. A random iteration method for decoding iterated system compression codes to generate an image by parallel processing, comprising the steps of:

receiving and storing a plurality of compression codes in a decoder;

determining a plurality n of initial points in a display memory;

randomly selecting n of the compression codes with replacement;

concurrently applying the transformation corresponding to each of the selected compression codes in one of a plurality of processing units to a particular initial point selected from the plurality of initial points without replacement to generate a set of n transformed points;

repeatedly iterating by carrying out the above steps of randomly selecting n compression codes and applying the corresponding transformations, wherein for each iteration, the set of n transformed points generated in the previous iteration is utilized as the plurality of n initial points for the following iteration, and continuing such iteration until an attractor image stabilizes; and

displaying data corresponding to the attractor as decoded output image.

61. A random iteration method for decoding iterated system compression codes to generate an image by parallel processing, comprising the steps of:

receiving and storing a plurality of compression codes in a decoder;

independently and concurrently, with a plurality of n processors, and until an attractor image stabilizes, carrying out the following steps of;

determining an initial point in a display memory,

randomly selecting one of the plurality of compression codes,

applying the transformation corresponding to the selected compression code to the initial point to generate a transformed point, and

repeatedly iterating by carrying out the above steps of selecting a compression code and applying the corresponding transformation, wherein for each iteration, the transformed point generated in the previous iteration is utilized as the initial point of the following iteration; and

after the attractor image stabilizes, displaying data corresponding to the attractor as a decoded output image.