Research Of Multispectral And Panchromatic Image Fusion Algorithms Based On Multi-scale Analysis

Image fusion combines two different types of images collected from the same scene to extract their own valuable information respectively,and finally synthesizes a process that contains more precise,richer and more abundant information.At present,image fusion has been applied in some other fields such as medical diagnosis,object detection,military reconnaissance and so on,therefore it has attracted such attention from researchers.Image fusion can be divided into infrared and visible image fusion,medical image fusion,multispectral and panchromatic image fusion and multi-focus image fusion according to the image attributes or imaging methods.Image fusion can be divided into pixel-level integration,feature-level integration and decision-level integration by fusion level.This paper mainly focuses on the pixel-level fusion of multispectral and panchromatic images,focusing on improving the image fusion efficiency and enhancing the image fusion quality.According to different research emphases,the paper deeply and systematically explored the feature extraction and fusion rules of images.The specific work arrangements of this paper are as follows:First of all,the background,significance and current situation of the research are introduced.The existing problems in the field of image fusion are summarized,and the evaluation criteria of image fusion quality are briefly introduced.This paper explores the classical fused image method,mainly discusses two kinds of fusion algorithms based on component replacement and multi-scale analysis,from IHS transform to PCA transform,from wavelet transform to contourlet transform,and analyzes the advantages and disadvantages of the two types of algorithms,the experimental simulation analysis of each algorithm.Next,aiming at the shortcomings of traditional nonsubsampled contourlet transform(NSCT)in image fusion,which is computationally intensive and inefficient,a fast multispectral and panchromatic image fusion algorithm is proposed.The paper mainly improved the structure of NSCT using multi-channel structure filter bank instead of the original tree structure filter bank.According to the characteristics of multispectral and panchromatic image data sources,combined with IHS transform,the algorithm efficiency is doubled without affecting the quality of the fused image.The experimental results show that the fusion image quality obtained by this algorithm is almost the same as the fusion image quality obtained by the traditional NSCT algorithm,and the computational time of the algorithm is greatly reduced.Finally,in order to obtain multi-spectral images with both rich spectral information and high spatial resolution,an image fusion algorithm based on fast nonsubsampled contourlet transform(FNSCT)and firefly optimization adaptive impulsive coupled neural network(PCNN)is proposed.Since PCNN algorithm parameters are often set by experience,the algorithm adapts the link strength and link range in PCNN,and introduces the firefly optimization algorithm with higher precision to optimize the other parameters of PCNN.Multiple experiments show that this method can achieve better objective indicators and subjective visual judgments in the fusion of multispectral and panchromatic images,and can improve the spatial resolution of multispectral images with higher spectral fidelity.