A Scene Matching Approach Based On Edge Signal Between IR And Visible Images

Some scene matching approaches between infrared and visible images are presented in this paper, in which there are imaging theorems, pre-processing method, similarity measures and matching strategies to be proposed. By the stable performances in the matching of multi-spectral images, Normalized Cross Gray-Correlation based on edges is chosen as our means.In the first chapter, the imaging models of infrared and visible images are given respectively, the similarities of the two spectrums are presented, which are the basic theorems of our approaches we proposed.In the second chapter, compared with different image pre-processing methods, an edge enhancement before histogram equalization is adopted in order to increase valid edge information and intensify the common characteristics between infrared image and visible image. The experimental results show the validity of the proposed approach.In the third chapter, some usual similarity measures are discussed, such as classical Cross Gray-Correlation, Mutual Information, Phase Correlation, etc. With some criteria as the matching probability, edge density and signal noise ratio, the differences between Space Domain and Frequency, correlation coefficient and mutual information are fully described.In the last chapter, several practical matching strategies are used to improve probability, precision and robustness of the matching system. To accelerate the matching speed, the SSDA reduces the computing amount, while multi-resolution theory used image pyramids decreases the number of seeking points, which is more efficient. Meanwhile, to raise the matching probability greatly, the matching strategy based on serial images is proposed.