Perceptual Edges Preservation conformal to Human Vision Perception

This paper presents a novel approach called PEPA (Perceptual Edges Preservation Algorithm) which enables computing machines to mimic the human vision capability in perceiving meaningful objects in an input scene, characterized in that the preserved edges are conformal to human vision perception. The approach mainly comprises three stages: (1) applying linear and nonlinear filtering to mimic the capability of human vision system in perceiving image objects, and (2) an Entropy-based energy function is proposed to overcome the gradient direction distortion due to sampling. (3) A Bayesian-based decision rule is developed to determine whether a target pixel is located on a perceptual edge. Simulation results are provided to justify the capability of imitating human visual perception.