一個解決色彩量化問題的幾何學方法

Abstract

在全彩色彩系統中，每個像素是用三個位元組來表示它的顏色， 一千六 百多萬種顏色可使用。然而在大部份的影像並不需要這 在此情況下，使 用全彩系統是浪費而且是不需要的。彩色銗X以較少的顏色來展現一張原 是用全彩色彩系統h是用統計方法來分群，在本篇論文中，我們提出一偽 臕扛熒s方法。運用幾何學中，一集合內兩點之間磳僄荈隻X越緊密的特性 來做分群。我們提出的方法h而且它的結果並不比其它方法差，足以應付 即時系 In RGB true color system, we use 3 bytes to represent the color of each pixel and there are 16777216 different colors in the system. But most images contain only a small subset of the sixteen million colors. It is enough to represent those images by small number of colors and using true color space is expensive and unnecessary. The process of selecting a small number of representative colors from an image of higher color resolution is called color image quantization. Previous algorithm almost use statistics techniques, it need more computational time. In this thesis, a new and efficient color quantization algorithm based on computational geometry has been proposed. By geometric computing characteristic, the shorter of diameter in a set means the closer of the set, and this is our basis for finding clusters. Our proposed algorithm is very fast than other methods and the quality is not worst than other methods, it is suitable for real time system.