Multiscale fusion of wavelet-domain information and clustering analysis for digital halftoning

A novel approach to improve the halftoning image quality by a mixture distortion criterion is the combination of a edge weighted least squares depending on the fusion multiscale information and the region weighted least squares depending on the improved K-means clustering method. The multiscale characterization of the original image using the discrete wavelet transform is obtained. The boundary information of the target image is fused by the wavelet coefficients of the correlation between wavelet transform layers, to increase the pixel resolution scale. The inter-scale fusion method to gain fusion coefficient of the fine-scale is applied, which takes into account the detail of the image and approximate information, where the fusion coefficient is referred to as the weighting operator, and to establish the boundary energy function. The improved K-means clustering method is used to segment an image several regions and 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 to establish the region energy function. In the halftone process, each clustering uses the weighted least-squares method through energy minimization via direct binary search algorithm, to gain halftoning image. Simulation results on typical test images further confirm the performance of the new approach.