A Research On Image Enhancement Algorithm Based On The Fusion Mechanism Of Biological Vision

Many creatures evolved a set of precise visual system which can integrate a variety of complex external optical signals through long-term training and adaptation to the complex and changeable natural environment for their own survival and predation. Through the study of the principles and characteristics of the biological vision we can be inspired to apply some knowledge of biological vision to solve present image problems. This paper bases on the mechanism of the rattle snakes' dual mode fusion cells and the human retina characteristics of adaptively enhancing brightness to respectively solve the problem of the fusion of visible light images and infrared images and the problem of high dynamic image display. The two aspects of contents of this paper are showing as follows:1. The rattle snakes can detect infrared signals and optical signals which will be fused together in the optic tectum at the head area. The experiments suggest that there are six kinds of fusion cells. According to the different characteristics of the six kinds of fusion cells on image processing, we analyze which is the most fusion mechanism of the six for gray images fusion and propose an improved false-color images fusion method based on the method proposed by A. Waxman et al. For our proposed method, the infrared target without optical information can be enhanced compared with A. Waxman's.2. The human visual system has a powerful adaptive regulation for external illumination. The cone and rod cells can process the brightness adaptively. The bipolar cells and ganglion cells can enhance pixels in the receptive fields locally. So information loss of high dynamic image display can be solved according to the characteristics of human visual adaptive ability. Based on human visual theory we proposed a fusion algorithm with global and local process and color correction. The subjective and objective evaluations of the new algorithm suggest that the method is better than most previous classics high dynamic image enhancement algorithms comprehensively.