多进制小波理论在SPOT和TM影像融合中的应用

Fusion of SPOT-P and TM Multispectral Image by M-band Wavelet Theory

二进制小波不适合融合SPOT全色影像和TM多光谱影像 ;三进制下的Mallat算法对SPOT全色影像和TM多光谱影像进行融合时 ,由于存在抽取和插值运算 ,使得融合的结果存在相位失真。为了克服相位失真 ,本文采用了多进制下的MRAGM算法。试验表明 ,它能有效地用于融合SPOT全色影像和TM多光谱影像 ,实现它的关键是构建具有紧支撑特性的M进制低通尺度函数滤波器。

Image fusion algorithms have evolved from traditional image fusion method,pyramid based method to nowadays wavelet based image fusion methods.The current wavelet based fusion algorithms are based on Mallat algorithm and à trous algorithm,and it deals with dyadic wavelet theory.The fundamental principle of those algorithms are: in frequency domain,substituting high resolution panchromatic image's low pass half band frequency with TM multi_spectral image's frequency,and then using invert Fourier transform to get fused image.But as the resolution ratio between SPOT PAN and TM image is 1∶3,it is not proper to use dyadic wavelet to solve this fusion problem.Zhu C Q has reported his research which is based on M_band theory,but M_band Mallat algorithm will lead to phase distortion in image fusion process.Wang Z J(2000) introduced MRAGM algorithm which is also based on M_band theory and can deal with any integer ratio of two images.This paper explains how to use MRAGM algorithm to fuse SPOT_P and TM_M.The result shows MRAGM improves the spatial resolution of multi_spectral images while keeps the hue unchanged.The key problem to use this algorithm is to designe compactly supported K_regularity M_band scale function filter.