Doctor of Philosophy (Ph.D.)
The presentation of color images on devices with limited color capabilities requires a reduction in the number of colors contained in the images. Color image quantization is the process of reducing the number of colors used in an image while maintaining its appearance as much as possible. This reduction is performed using a color image quantization algorithm. The quantization algorithm attempts to select k colors that best represent the contents of the image. The original image is then recolored using the representative colors. to improve the resulting image, a dithering process can be used in place of the recoloring.;This dissertation deals with several areas of the color image quantization process. The main objective, however, is new or improved algorithms for the production of images with a better visual quality than those produced by existing algorithms while maintaining approximately the same running time. First, a new algorithm is developed for the selection of the representative color set. The results produced by the new algorithm are better both visually and quantitatively when compared to existing algorithms. Second, a new nearest-neighbor search algorithm that is based on the Locally Sorted Search algorithm is developed to reduce the time required to map the input colors to a representative color. Finally, two modifications are made to the error-diffusion dithering technique that improve the execution time. These modifications include the use of a two-weight matrix for the distribution of the error values and the presentation of a method to parallelize the error-diffusion technique. Furthermore, the analytical results of several experiments are provided to show the effectiveness each of these additions and improvements.
© The Author
Necaise, Rance David, "Improvements to the color quantization process" (1998). Dissertations, Theses, and Masters Projects. Paper 1539623933.