Visual Information Evaluation With Entropy of Primitive
Songchao Tan, Shurun Wang, Xiang Zhang, Shanshe Wang, Shiqi Wang, Siwei Ma, Wen Gao · IEEE Access · 2018
In this paper, we overview the recent work on entropy of primitive (EoP), including its concept, design, extension, and mathematical analysis in evaluating the visual information of natural images. The design philosophy of EoP is establishing an entropy model that quantifies the visual information based on patch-level sparse representation, due to the close relationship between sparse representation and the hierarchical cognitive process of human perception. Furthermore, based on the concept and definition of EoP, we also demonstrate several applications, including just noticeable difference estimation and visual quality assessment. The future research directions of visual information evaluation are also envisioned, where we can perceive both promises and challenges.