Study On The Fusing And De-Noising Method Based On Wavelet Transform For Remote Sensing Images

The multi—spectral image of Landsat TM has been concerned and applied for its plentiful multi—spectral imformation and relative low price.The method of fusing and de-noising for Landsat TM image is hot and important in the remote research realm.In recent years,some methods of fusing and de-noising for multi—spectral image of Landsat TM is popular,such as,image lowpass filtering based on Fourier transform in frequency domain,image median filtering or adaptive filtering in spatial domain,PCA transform fusion,HIS transform fusion, Wavelet transform fusion,Brovey transform fusion,Pansharp transform fusion,Gramschmidt transform fusion,etc.Wavelet transform is local in the spatial domain frequency domain,it can get low or high frequency character in image singal freely and easely,and the low or high frequency character is a adaptive selection for requirement.So,in recent years,Wavelet transform is widely used in image processing field.In this paper,de-noising method of soft-threshold based on Wavelet Transform is adopted to remove white noise and little fragmentary plot from multi——spectral image of Landsat TM;image fusion methods of HIS+DWT and PCA+DWT are adopted to fuse multi—spectral image of Landsat and PAN of SPOT,PAN of SPOT has high space resolution of 10m.Evaluate standards for de-noised and fused multi—spectral image ofLandsat TM include statistical criterion and visua 1.criterionThe following is main points in this paper:(1).The image configuration based on wavelet muli_resolution decomposition is similar to"pyramid".So,it is easy to get low frequency character or high frequency character(horizon,vertical,diagonal)in image singal freely and easely.Therefore, general picture information(low frequency character)in the remote sensing image with low space revolution is extracted successfully;ground detail information(high frequency character)in the remote sensing images with low space revolution is extract successfully too.(2).The de-noising method of soft—threshold based on Wavelet Transform is adopted to remove white noise and little fragmentary plot from multi—spectral image of Landsat TM.And,ground detail information(high frequency character,such as rivers,roads,edges of targets)is preserved effectively in denoised image.So it is possible to carry out other further classification for multi—spectral image of Landsat TM,such as,supervised classification,sub-region classification,etc.(3).With the fusion method ofPCA+DWT,the fused multi—spectral image of Landsat TM preserves its special information gmostly,at the same time,the ground detail information in the SPOT PAN is fused,greatly.(4).With the fusion method of IHS+DWT,ether under TM432 or TM453 pseudo color mode,the fused multi—spectral image of Landsat TM preserve special information mostly, at the same time,its space resolution is raised greatly.So both special and space revolution of fused image is higher.The result is satisfied.