Research On Theory And Method Of Multi-source Image Fusion

Muti-source image fusion is a technique that the original images will be processedbased on some algorithm to obtain a new one, a new image to satisfy some requirements.This paper mainly studies remote sensing image fusion algorithm, which are dividedinto conventional fusion algorithm and fusion algorithm based on multi-resolutionanalysis. First, the conventional algorithms are introduced, such as: componentreplacement algorithm (HIS, PCA), Brovey and high-pass filtering algorithm. Thesubjective and objective evaluation criteria are discussed, this paper elaborates thefundamental principles of13kinds of objective evaluation criteria. Second, this paperfocuses on the Laplacian Pyramid Decomposition, Wavelet Transform, CurveletTransform and Nonsubsampled Contourlet Transform (NSCT) et al multi-resolutionanalytical tools for image fusion and the effect of decomposition layers on fusionperformance is studied. In view of the good nature of the NSCT in image decomposition,the new fusion algorithm based on Nonsubsampled Contourlet transform deep-leveldecomposition is proposed. The fusion experiments show that the fusion performance isimproved, according to the resolution of the images involved in the integration ofdifferent layers NSCT decomposition. Last, this paper describes the design of imagefusion software. All fusion algorithms are integrated in the software, which has goodman-machine interactive interface software. Algorithms and software performance isstable and reliable operation.