Effect of Wavelet Basis and Decomposition Levels on Performance of Fusion Images from Remotely Sensed Data
陈健飞,龚建周,夏北成,刘彦随
2010-01-01
Abstract:The integration of wavelet transform with IHS has apparently become popular to incorporate multi-sources remotely sensed data to create a new image,in that better fusion results might be produced by the method.But related factors are poorly considered,which would directly affect the fusion results,such as wavelet basis,decomposition levels.In Matlab,SPOT images,including panchromatic band and three multispectral bands,were used as analytical data here.Fusion images were created by highlighting those factors:different cluster of wavelets,wavelet basis with different serial number,and decomposition levels.Three indices for image performance,including entropy (joint entropy),average gradient and deviation,were calculated.All fusion images had similar performance with only one wavelet decomposition level.But fusion images had different performance with increasing levels and different wavelet basis.Three wavelet basis,including coif5,sym5,dmey,did not show different performance with changing levels.Two wavelet basis,db1 and bior3.1,had unique response characteristic within different ranges of level,such as indices value were stable while levels between 1 and 4,and then monotonically decreasing.Taking rbio3.1 as wavelet basis,the distortion image was created when the wavelet decomposition levels was equal to 4.With the increasing levels,fusion images were created and characterized by dramatic distortion in image spectral.Besides,different cluster of wavelets in Matlab has different fusion image in performance.The paper could better support the practical application of remote sensed data in defining method to improve image performance by fusion multi-sources data and make the operation simpler and faster.