Context sensitive color quantization system and method

Claim 1

A color image quantization system useable in producing a second image limited to a number of selected colors from a first image generally having a number of colors greater than the number of the selected colors, said color image reconstruction system comprising: first image input means for digitally providing first image data representative of respective colors at a multiplicity of individual pixels in the first image; input buffer means in operative communication with said first image input means for temporarily storing the first image data; image quantization means in operative communication with said input buffer means for receiving the first image data and determining the selected colors to be included in the second image, said image quantization means further determining which of the selected colors respectively correspond to individual pixels of the second image; said image quantization means further operative at least in part to selectively account for perceptive blending of at least two different colors appearing in the first image in determining the selected colors to be included in the second image, based at least in part on respective locations of the individual pixels of the first image in which the two different colors appear; wherein said image quantization means includes Hopfield neural network means for determining the selected colors to be included in the second image; and second image output means in operative communication with said image quantization means for digitally outputting second image data representative of respective colors in the individual pixels of the second image.

Claim 15

A method of producing a second image limited to a number of selected colors from a first image generally having a number of colors greater than the number of the selected colors, said method comprising the steps of: (a) reading digital data representative of respective colors at a multiplicity of individual pixels of the first image; (b) determining the selected colors to be included in the second image based on the respective colors at the multiplicity of individual pixels of the first image; (c) determining which of the selected colors respectively correspond to individual pixels of the second image; (d) determining the selected colors respectively corresponding to the individual pixels of the second image in step (c) by selectively determining which colors of the selected colors should respectively correspond to adjacent pixels of the second image to substantially approximate the perceptive blending of respective colors in corresponding pixels of the first image; (e) wherein steps (b), (c) and (d) collectively include the following steps: (f) defining a color set of possible colors including first colors in the first image and second colors representing the perceptive blending of two different colors which appear in adjacent pixels in the first image; (g) defining a first tag function for respective colors in the color set, said first tag function weighted according to prevalence of occurrence of the respective color in the first image and by a measure indicative of a desirability of utilizing the respective color to replace two other colors occurring adjacently in the first image; (h) defining a second tag function for selected pairs of colors in the color set, said second tag function weighted to facilitate choosing of only one color of each selected pair of colors as colors of the respective selected pair become more similar and further weighted according to prevalence of adjacent occurrence of one color of each selected pair and another color in the first image which average to another color of each selected pair; (i) determining a table of the selected colors from possible combinations of colors within the color set based on an optimization of a selection function defined as a sum of the first and second tag functions, wherein the optimization of the selection function is performed in step (i) by employing a Hopfield neural network; and (j) mapping colors from the table of the selected colors to individual pixels of the second image by selectively determining which colors of the selected colors should respectively correspond to adjacent of the individual pixels in the second image to approximate the perceptive blending of respective colors in corresponding pixels in the first image; and (k) digitally outputting second image data representative of the respective colors in the individual pixels of the second image.

Claim 16

A method of producing a second image limited to a number of selected colors from a first image generally having a number of colors greater than the number of the selected colors, said method comprising the steps of: (a) reading digital data representative of respective colors at a multiplicity of individual pixels of the first image; (b) determining the selected colors to be included in the second image based on the respective colors at the multiplicity of individual pixels of the first image; (c) determining which of the selected colors respectively correspond to individual pixels of the second image; (d) determining the selected colors respectively corresponding to the individual pixels of the second image in step (c) by selectively determining which colors of the selected colors should respectively correspond to adjacent pixels of the second image to substantially approximate the perceptive blending of respective colors in corresponding pixels of the first image; (e) wherein steps (b), (c) and (d) collectively include the following steps: (f) defining a color set of possible colors including first colors in the first image and second colors representing the perceptive blending of two different colors which appear in adjacent pixels in the first image; (g) defining a first tag function for respective colors in the color set, said first tag function weighted according to prevalence of occurrence of the respective color in the first image and by a measure indicative of a desirability of utilizing the respective color to replace two other colors occurring adjacently in the first image; (h) defining a second tag function for selected pairs of colors in the color set, said second tag function weighted to facilitate choosing of only one color of each selected pair of colors as colors of the respective selected pair become more similar and further weighted according to prevalence of adjacent occurrence of one color of each selected pair and another color in the first image which average to another color of each selected pair; (i) determining a table of the selected colors from possible combinations of colors within the color set based on an optimization of a selection function defined as a sum of the first and second tag functions, wherein the optimization of the selection function is performed in step (i) by employing a Hopfield neural network; and (j) mapping colors from the table of the selected colors to individual pixels of the second image by selectively determining which colors of the selected colors should respectively correspond to adjacent of the individual pixels in the second image to approximate the perceptive blending of respective colors in corresponding pixels in the first image; and (k) digitally outputting second image data representative of the respective colors in the individual pixels of the second image.