Image Quality Metrics Based on Single and Multi-Channel Models of Visual Processing

We review two classes of image analysis tools based on single and multiple channel models of human vision processing, designed to predict the visibility of printed dots and halftone texture, respectively. 1 Introduction Nearly every imaging peripheral Hewlett-Packard manufactures is designed to optimize subjective image quality. Whether we are capturing, transmitting or rendering visual images, we consider the impact that our design decisions have upon subjective image quality. Subjective impressions of image quality, however, are difficult, if not impossible, to measure, quantify and predict. Subjective image quality is influenced by personal experience and other high-level factors outside of any simple physical description of the image. Thus we are faced with the challenging task of optimizing a signal that we find difficult to measure. Some physical attributes of an image can be strongly correlated with perceived quality. Our approach to image quality evaluation is to identify the..