A Realistic Image Rendition Method Based on the Cone Adaptation Model
一种基于视锥适应模型的真实影像再现方法

A cone adaptation model based method for realistic image rendition is proposed after analysis of the disadvantages of several tone rendition algorithms.In this method,the half-saturation constant of the original cone cell transduction function varies with input intensity,while the ratio between the half-saturation constant and the half-pigment bleach constant of the function is kept invariant to be in accordance with the effects of pigment bleaching and neural feedback on cone adaptation.Many simulation results and comparison between this method and some other tone reproduction algorithms show that this method can illustrate good performance of dynamic range compression while keeping color constancy,greatly reduce the "halo" artifacts frequently occurring in the field of tone reproduction,and avoid the "graying out" effects brought about by models such as the retinex and the opponent neural network.Since the convolution kernel used for computation of the ambient intensity is of very small size,its computational speed is far exceeding that of some classical image rendition methods and it may be used in some real-time applications.