Search

Curvelet transform,as a method of multiscale geometric analysis,is more suitable for image processing than wavelet,and more appropriate for analyzing the image edge characteristics of curve and line,which has better approximation precision and sparsity description.In this paper,the methods of integrating SAR image and MS image are proposed based on curvelet transform.SAR image and I image,which is given by a linear HSI transform are given by curvelet transform to obtain coarse coefficients and detail coefficients.The new coarse coefficients are obtained by using edge information and features of SAR image.The detail coefficients are dealt with traditional method that gets max of detail coefficients both I and SAR image.Then,the inverse curvelet transform gets the new intensity I image.Finally degree of distortion and space frequency are used to evaluate the result.The results of experiment indicate that the method excels those of based on HSI or curvelet transform in preserving spectral information and enhancing resolution.

Applying the morphological operator,which characterizes the spatial correlative informations of pixels,to endmember extraction of hyperspectral remote sensing image can improve the performance of algorithm effectively.In order to overcome the limitations in sorting rules and replacing criteria of extended-morphological operator,which is commonly used in hyperspectral remote sensing image to extract endmembers,the modified extended-morphological operator is proposed after introducing the concept and presenting the calculating method of reference vector.The cross-replacement phenomena at the junction of different classes can be avoided and the correct coverage direction can be ensured when the modified sorting rules and replacing criteria have been applied in endmember extraction algorithm to enhance the results as key means.The endmember extraction algorithm using the determine profiles,generated after open-close and close-open operations defined by basic dilation and erosion operations of modified extended morphology,is described in detail.The automated modified extended-morphological endmember extraction algorithm is achieved by using both spatial and spectral information in a combined manner,thus,the endmember extraction result is superior to the approachs designed from a spectral information viewpoint only.The alogrithm is implemented in IDL7.0 and testd by using real hyperspectral imagery collected by airborne visible/infrared imaging spectrometer in cuprite area,the experimental results of the similarity on spectral curves,the average similarity and the mineral distribution maps verified the validity of the algorithm.

Dimensionality reduction is a common preprocessing for hyperspectral image,and Kernel Principal Components Analysis(KPCA),as a common feature extraction method,makes use of nonlinear mapping to capture higher-order statistics of data.An optimization sample set algorithm,which is used in KPCA for dimensionality reduction of hyperspectral image was proposed.This algorithm picks sample set used in KPCA taking the statistics of the whole hyperspectral image into account simultaneously,and the minimum sample set with similar energy distribution of the full image is the final selection.The algorithm was implemented in IDL7.0 and tested by using the real hyperspectral image from Cuprite.The experiment results show that the new algorithm is able to save computing time significantly and perform better than conventional KPCA in dimensionality reduction.

A novel image fusion method based on Contourlet domain hidden Markov tree models is proposed.Contourlet transform provides a flexible multiresolution,local and directional image expansion,and also a sparse representation for two-dimensional piecewise smooth signals building images.Contourlet HMT can capture all inter-scale,inter-direction,and inter-location dependencies of the Contourlet coefficients.Aiming at the different frequency bands of Contourlet decomposition with different characteristics,different fusion rules are applied to different subbands.In the low-frequency information,the weighted average mean is used to obtain the fused low-frequency information.Contourlet HMT is applied to design low-frequency information rule,the fusion method has the ability to strengthen the relationship among the Contourlet coefficients,extract more detailed and exact information from the original images.The fused images by the proposed algorithm exhibit good performance both in subjective and objective standards.Experimental results also show the simplicity and effectiveness of the method and its advantages over the conventional approaches.