Detail selection incorporating subjective factors for very low bit-rate image coding

This paper is concerned with the subjective selection of image details for segmentation-based very low bit-rate coding. An approach is proposed for extracting image details and ranking the details in the order of perceptual significance to the human visual system (HVS). A perceptual ranking model has been established by multivariate regression analysis based on the data obtained from subjective tests. In terms of the correlation between the objective ranking and subjective ranking of perceptual significance and subjective image quality, this model provides better results in detail ranking than that provided by an empirical ranking formula proposed in a previous study. Segmented regions and selected details are coded to illustrate the efficiency of coding meaningful details. Compared with the pure segmented images, the addition of the selected details improves the subjective image quality at lower bit rates