Color printing image coding using vector quantization

Abstract

A color printing image coding method is proposed. In this method, four colors (cyan, magenta, yellow, and black) are used to form a four-dimensional vector. The image is partitioned into small blocks, and a codebook is updated from block to block. The codebook size is adaptively determined by the local statistics, i.e., the variance, range, and summation of the ranges. These three determination methods are compared by computer simulation, and the results indicate that the best performance was obtained with summation of the ranges. Further simulation was performed and excellent image quality was achieved. Additionally, the present method was improved on in two areas: to reduce computation requirements as splitting algorithm was used, and then variable length coding (either run length coding or quad tree expression) was applied to this algorithm to reduce spacial redundancy of quantized code, thereby obtaining a gain of 5–20 dB in signal-to-noise ratio.