Wc optimizc thc noisc shaping bchavior of color crror diffusion by designing an optirnizcd crror filtcr bascd on a proposed noisc shaping modcl for color crror diffusion and a gcncralizcd lincar spatially-invariant modcl of thc human visual systcm. Our approach allows thc crror filtcr to havc matrix-valued cocfficients and diffusc quantization crror across channcls in an opponcnt color rcprcscntation. Thus, thc noisc is shaped into frcqucncy rcgions of rcduccd human color scnsitivity. To obtain thc optimal filtcr, wc dcrivc a matrix version of the Yulc-Walkcrcquations which wc solvc by using a gradicnt dcsccnt algorithm. Brian L. Evans Embedded Signal Processing Laboratory Dept. of Electrical and Computer Engineering The University of Texas at Austin Austin, TX 78712-1084 E-mail: bevansQece.utexas.edu In this paper, wc dcrivc the optimum matrix-valued crror filtcr using thc matrix gain modcl [3] to modcl thc noisc shaping bchavior of color error diffusion and a gcneralizcd lincar spat...
Niranjan Damera-Venkata, Brian L. Evans