Digital Halftoning Based on Clustering Analysis and Weighted Least Squares

We propose a weighted least-squares-based halftoning model from human vision system(HVS) model and an efficient iterative strategy using gray image statistical information. First, the gray image is partitioned into several finite regions using clustering segmentation method. Next, the statistics of the mean and variance of the gray-scale pixel of each clustering are calculated. Finally, the new energy function is constructed using the weighted least squares method, which the reciprocal of the variance of the segmented regions are referred to as the weighting operator. Our method also incorporates a measurement based printer dot interaction model to prevent the artifacts due to dot overlap and to improve texture quality. The analysis and simulation results show that the proposed algorithm produces better gray-scale halftone image quality when we increase the number of clustering with a certain range. A performance measure for halftone images is used to evaluate our algorithm. The value of MSEv and PSNR for the partitions regions that the proposed algorithm achieves consistently better values of MSEv and PSNR than the LSMB algorithm. After four iterations of the algorithm of the proposed algorithm, the convergence error dropped to 0.25.