Comparision Of Multiscale Transform Fusion Methods For Multiband Image

In order to make full use of diversified data sources,it is necessary to effectively combine sensors with different characteristics to obtain more complete and clear images,which is more suitable for human eye vision,computer classification,and recognition.Aiming at the current research on image fusion focuses on two images and cannot meet the needs of multi-band detection and development,this paper compares the common multi-scale transformation methods and fusion rules,generalization to multi-band image fusion.The main work is as follows:(1)A comparison of multi-scale image fusion based on multi-scale transform.In order to select a multi-scale transform method with better fusion effect to improve the quality of multiband image fusion,and the simulation experiments of Pyramid transform,wavelet transform and series of methods based on multi-scale geometric analysis theory were conducted under the same fusion on the basis of analyzing the characteristics of different multi-scale transformation methods,and analyze and evaluate the results of the fusion from the perspective of the subjective and the objective to select the best multiscale transformation method.(2)A comparison of Multiscale Transform Fusion Rules..In order to select better fusion rules to improve the quality of multi-band image fusion,multi-band image fusion is performed on the best fusion result chosen in work(1),using the method of taking the maximum value of absolute value,consistency tests,taking the maximum value of regional energy,taking the maximum of regional variance in the high frequency part and weighted average,Gaussian fuzzy logic,and area matching at low frequencies,and select the fusion rules with better fusion results by subjective and objective evaluation.(3)Design and Implementation of Multiband Image Fusion System.The entire fusion system was developed and designed using matlab,which realized the fusion of multi-band images with different high-and low-frequency fusion rules in different multi-scale transform methods.