An efficient algorithm for color quantization based on principal component analysis and dynamic programming is described. The purpose of this paper was to summarize work done in the field and present the algorithm in the context of what has recently been done. This paper includes previous work, a description of the new algorithm, complexity analysis, and some examples of its use. The length of the paper is suitable and it appears to fulfill its intended purpose.

The best features of the paper are the problem formulation and literature survey. I found the mathematics to be presented poorly. Overall, this paper is somewhat difficult to read and contains too much detail for a large part of the intended audience. Since the intention of the paper was to present the algorithm, the analysis included was appropriate, but the author would do well to present the material again with less detail for the larger audience. To read the current version requires an understanding of algorithm analysis, linear algebra, and graphics and image processing. The main idea can be conveyed with less detail.

This paper is harder to read than the two papers that it was compared to, describing the median cut [1] and minimum variance [2] methods. The references provided are adequate to learn about this area of research. Unfortunately, this paper relies too heavily on the reader’s ability to abstract concepts that could be easily conveyed with a few simple diagrams. The experiments presented are sparse, heavily concentrating on one of the two images given as raw data. I was left with more questions than answers.