On The Color Quantization Algorithm Based On Vision Properties And Its Application

The Computer Assistant Design System is widely applying to the industry fields of carpet, weaving, printing and so on. Acquiring the most representative colors information is one of the pivotal technology on the drafts design and the stuff quantification, such as the various color yarn quantification of a carpet designation.Now,most of the Computer Drafts Design Assistant Systems have some drawbacks such as incomplete color information,inconsistent systems , man-machine interaction, inaccurate center colors,slow computing speed,and so on.And most of the given techniques cannot apply directly to the industry stuff design. Because the color information acquired exactly from a image must take synthetically account of the least distortion, arithmetic complexity, human vision feature and so on, there is scarely a satisfying algorithm which can completely meet with color layers and some essential details at present.By analysizing of the given techniques for color information acquiring and the existent quantization algorithms and image data in the draft design system, we propose a new algorithm for color quantization based the vision feature in this paper. The differences of this algorithm from the existent quantization algorithms are that it congregates the merit of conventional algorithms for the color and space information of image. The idiographic implementation is that: it finds out the colors of the leading image style as one part of the initial clustering centers firstly, then turns up the colors which have well characters from other colors and makes these colors as other part of the initial clustering centers, and locates dynamicly out the best clustering centers according to the rule of least comparability with clustering center colors during the process of clustering the remnant color samples, lastly makes the rebuilt image by those clustering center colors as the representative colors.The color comparability considers not only color's frequency but also chromastism and evaluates some weight. The algorithm can correspond with the contradiction between color layers and essential details by adjusting the weights of color comparability and make the color quantization as optimal as possible.We have performed the novel algorithm in VC++ environment. We fulfill the experiments by various selected images, and compare the algorithm with other on the aspects of experiment effects, mean distortion and space-time complexity. The experiments show that the algorithm is effective and has higher quantization performance which can adjust the emphatic degree of color layers and essential details by modifying the weights to meet various quantization tasks.