摘要：Color quantization is essential due to the limitations of image displays, data storage, and data transmission. The process of color image quantization can basically be divided into two major steps: color palette design and pixel mapping. Many algorithms have been proposed for the design of the color palette and for pixel mapping. Among them, PMRC is a fast pixel mapping algorithm, which uses the concept of reduction of color space dimensionality [1]. Experimental results have shown that PMRC is much faster than the traditional exhaustive search method. However, there is room for improvement. In this paper, a new pixel mapping method for color quantization using principal component analysis is proposed. Our proposed method first uses the first principal component axis to replace the projection line in PMRC. Next, a new projection value search algorithm is used to solve the shortcomings of PMRC. Finally, a three-dimensional projection filter method is employed to further accelerate the search speed. The experimental results show that the execution speed of our proposed method is much faster than that of the exhaustive search method. Moreover, our proposed algorithm achieves the time complexity of O(NK), which is superior to the O(NKlogK) of PMRC. Our proposed method is therefore a very efficient method for pixel mapping in color quantization techniques.
关键词：color image quantization; pixel mapping; principal component analysis
