Color image quantization using a hierarchical color perception model
AssigneeMicrosoft Corporation
InventorsHeung-Yeung Shum, Wei-Jun Chen
Life of this patent
Estimating the term…
Abstract
Color image quantization using a hierarchical perceptual color model is disclosed. In one embodiment, a method constructs a clustering space of an image having a number of pixels, based on a color perception model. The clustering space includes a number of significant pixels, where each of the significant pixels does not have a parent pixel within the clustering space. The colors of the image are quantized based on these significant pixels. The remaining pixels have their colors mapped to one of the quantized colors.
Claims16
A computer-implemented method comprising:
- constructing a clustering space of an image having a plurality of pixels based on a color perception model, the clustering space comprising a plurality of significant pixels selected from the plurality of pixels of the image, such that each of the pixels of the image other than the significant pixels has at least one parent pixel;
- quantizing colors of the image based on, for each significant pixel, the color of that significant pixel and the pixels of the image which have that significant pixel as a parent pixel;
- and,
- mapping each of the pixels of the image other than the significant pixels of the clustering space to one of the colors as quantized, selected from quantized colors of the at least one parent pixel of the pixel;
- wherein constructing the clustering space comprises,
- growing a perceptible color region of each pixel of the image, and,
- building a relationship graph of parent-child relationships of the pixels based on the perceptible color regions of the pixels of the image.
A computer-implemented method comprising:
- constructing a clustering space of an image having a plurality of pixels based on a color perception model, the clustering space comprising a plurality of significant pixels selected from the plurality of pixels of the image, such that each of the pixels of the image other than the significant pixels has at least one parent pixel;
- quantizing colors of the image based on, for each significant pixel, the color of that significant pixel and the pixels of the image which have that significant pixel as a parent pixel;
- and,
- mapping each of the pixels of the image other than the significant pixels of the clustering space to one of the colors as quantized, selected from quantized colors of the at least one parent pixel of the pixel;
- wherein quantizing colors of the image comprises,
- yielding cluster centroids of significant colors from colors of the significant pixels.
A machine-readable medium having instructions stored thereon for execution by a processor to perform a method comprising:
- growing a perceptible color disk of each of a plurality of pixels of an image based on a color perception model;
- building a relationship graph of parent-child relationships of the pixels based on the perceptible color disks of the pixels of the image;
- determining a clustering space comprising a plurality of significant pixels as the pixels of the image having no parent pixels within the relationship graph;
- quantizing colors of the image based on the significant pixels of the clustering space;
- and,
- mapping each of the pixels of the image other than the significant pixels of the clustering space to one of the colors as quantized, selected from quantized colors of at least one parent pixel of the pixel.
Dependencies are inferred from the claim text.