Multi-level Vector Error Diffusion for Smear Artifacts Reduction in Boundary Regions

©2004, IS&T—The Society for Imaging Science and Technology classified into scalar error diffusion methods and vector error diffusion methods depending on the correlation among the CMY color channels. For scalar error diffusion using CMY separation, only a monochromatic version is processed for each color channel in the color image. Thereafter, the halftoned channels are recombined according to a recombination rule to obtain the desired halftone. Multi-level scalar error diffusion is then essentially the same as bi-level scalar error diffusion, except that the thresholding unit is replaced with a quantizer. Yet, in this case, the dot patterns generated by the superposition of the three individually processed components are not controlled systematically, thus colorimetric color reproduction is not inherently guaranteed. Meanwhile, vector error diffusion is carried out in a device independent color space, such as CIEXYZ or CIELAB color space.10–12 The input RGB image is represented in a vector form composed of XYZ or LAB values, plus the known XYZ or LAB values of eight primary colors (white, cyan, magenta, yellow, red, green, blue, and black) are also used.12 Therefore, vector error diffusion generates a proper dot pattern based on a vector operation, leading to a colorimetrically correct image. Haneishi and co-workers12 suggested that vector error diffusion using CIELAB color space achieves the best image quality. However, vector error diffusion generates smear artifacts around boundary regions. These artifacts result from the formation of false edges in areas where there was originally a smooth transition between two colors; then the quantization error Introduction