Multi-Sensor Image Enhancement and Fusion for Vision Clarity Using Contourlet Transform

Abstract

A multi- sensor pixel based image fusion system using multi-resolution (MR) approach is proposed. Images captured by visual cameras are fused with those captured by night vision cameras for visibility improvement of a surveillance system and for driving at night or in bad weather conditions. Images captured by visual cameras in insufficient illumination generally have dark shadows and low contrast. This problem seriously affects other further processing schemes. Before fusion a pre visual image contrast enhancement process is done. Multi-resolution decomposition is performed at each input image and they are combined according to the predefined fusion rules. Different fusion rules are defined for the approximate and detail coefficients. The Contourlet transform is used for the multi-resolution decomposition. Contourlet transform has better performance in representing the image salient features such as edges, lines, curves and contours because of its anisotropy and directionality. It is therefore well-suited for multi-resolution image enhancement and fusion. The simulation results show that this approach gives encouraging results for applications in low light and/or non uniform lighting conditions. The composite image contains almost all of the salient features of the input images and it appears to be natural.