The Study of Realistic Image Rendition Algorithms Based on Human Visual System

Abstract

Realistic image rendition is to reproduce the human perception of natural scenes, where the "realistic" means providing machine vision with the ideal images according to human visual system. The main characteristics of realistic image rendition involve color constancy, lightness constancy, image enhancement and high dynamic range compression. In this paper, the developments on realistic image rendition are investigated, human visual system structure and characteristics are described in detail and some excellent algorithms are studied and discussed, especially the classical Retinex series algorithms and tone mapping algorithms. In the end, the performances of these algorithms are simulated and shown.

Implementation of realistic image rendition algorithm based on DSP

“…The problem of image capture and display devices' dynamic range does not match results in the loss of image information, especially in the dark areas and bright areas where details are difficult to distinguish. Therefore, it's necessary to compress the dynamic range of input image so as to show a colorful and detail-rich image [2] .…”