Fusion Infrared And Visible Images Based On Improved Multi-scale Analysis

With the rapid development of visible, infrared and other multi-sensor technology, image fusion plays a more important role in the field of computer vision, medical diagnostics, remote sensing, night reconnaissance, natural resource exploration and other fields. By integrating information from different bands, image fusion can not only avoid the redundancy of the image information from the multi-sensors, but also retain the complementary features among the source images. This efficient access to obtain information has a very important theoretical significance and broad application prospects both in the military field and in civilian industrial areas.This dissertation focuses on the integration of infrared and visible images using multi-scale analysis. Firstly, this dissertation summarize the relevant research results at home and abroad in the field of image fusion and furthermore introduce the image fusion classification system, the evaluation criteria system and classical image registration methods. Secondly, this dissertation select the multi-scale analysis of the pyramid decomposition and wavelet transform fusion as a basic tool for analysis and comparison. Through specific simulation and integration of comprehensive analysis, the results show that wavelet transform has a strong potential to enhance. Finally, based on the basic wavelet transform fusion method, this dissertation use a PCA-based wavelet low-frequency sub-band fusion method and integration with the region-energy-based wavelet high-frequency sub-band method. The enhanced method introduces PCA fusion rules to low-frequency sub-band and the region energy analysis method to the fusion rules of high frequency sub-band. Relevant simulation is made to compare with the previous methods. Simulation results show that the enhanced method has some improvement in the fusion effect such as standard deviation, mutual information and entropy compared with the previous methods.