The optimization of blue noise halftoning based on human visual perception

Halftoning is a technique to reproduce the appearance of continuous tone images with limited number of output levels. Among the many blue noise algorithms, the blue noise mask (BNM) was the first method to generate halftones with blue noise characteristics by a point process. It combines the advantages of the simplicity of point process and the visually pleasing characteristics of blue noise. The focus of this thesis is to optimize the BNM for color halftoning and multilevel halftoning. A human visual system model is incorporated into the halftone scheme design and halftone quality evaluation. Some psychological experiments are carried out to provide an experimental basis for the optimization.; A review of current digital halftoning techniques is presented, and the essential concepts and principles of blue noise halftoning are elaborated with the concentration on the digital filter techniques to construct BNM.; The combinational properties of blue noise binary patterns are analyzed. Based on the analysis, jointly-blue noise mask (JBNM) is constructed to generate a set of blue noise masks that produces high quality blue noise patterns whether the masks are used individually or jointly. The JBNM can be used in color halftoning to create visually pleasing images on both the single and combined planes.; The selection of intermediate output levels in multitoning is directly related to the quality of multitone reproduction. A psychological experiment is designed to investigate the effect of spatial modulation on lightness perception. An effective lightness space is developed based on the experimental results which provides a calibrated space to select the optimal output levels for multilevel printing.; A texture visibility metric is proposed which predicts the visibility of uniform color halftone patches at threshold level. This metric is represented by the critical viewing distance below which the visual error just exceeds the threshold defined by the de Vries-Rose law. An experiment is conducted to determine the texture visibility of color halftone patterns subjectively. The experimental results correlate well with the values predicted by the metric, indicating that the metric is able to predict the visibility of halftone patches over a wide range of texture characteristics.