A patch-based structural masking model with an application to compression

摘要: The ability of an image region to hide or mask a given target signal continues play key role in the design numerous processing and vision systems. However, current state-of-the-art models visual masking have been optimized for artificial targets placed upon unnatural backgrounds. In this paper, we (1) measure natural-image patches distortion; (2) analyze performance widely accepted standard model predicting these data; (3) report optimal parameters different patch types (textures, structures, edges). Our results reveal that does not generalize across type; rather, proper should be coupled with classification scheme which can adapt based on type content contained local patches. utility adaptive approach is demonstrated via spatially compression algorithm employs patch-based classification. Despite addition extra side information high degree spatial adaptivity, yields efficient wavelet strategy combined very accurate rate-control procedures.