Multi-Source Remote Sensing Image Fusion Algorithm Based On Shearlet Transform

Image fusion is the process that fuses multiple original images which obtained from different sensors into one image.With the development of the sensor technology,image fusion has been widely applied in the medical imaging,military and remote sensing fields.In this paper,according to the imaging characteristics of different types of remote sensing images,image fusion algorithms are proposed based on the theory of shearlet transform.According to the imaging characteristics of multi-spectral and panchromatic images,a remote sensing image fusion method based on Shearlet transform and region segmentation is proposed.In this algorithm,low frequency parts of multi-spectral image were segmented into regions by means of thresholds methods.To improve the edge blur of the fused image,the pair images' low frequency components were adopted weighted fusion method under the rules of the degree of region match(DRM).In high frequency components fusion,fusion rules changed to the ratio of region sharpness(RRS),in order to yield more detail information.The results show the proposed algorithm can not only retain spectrum information source by the multi-spectral image,but also improve the spatial detail information.In three main evaluation criteria of standard deviation,information entropy and spectral distortion,our algorithm has19.9% and 2.4% improvement,35.0% reduction compared with traditional nonsubsampled contourlet transform(NSCT)method,respectively.According to the imaging characteristics of SAR,infrared and visible images,by the study of PCNN,a fusion algorithm is presented based on nonsubsampled shearlet transform(NSST)and improved PCNN.In this algorithm,for the low frequency subbands,the output amplitude of the model is calculated by a sigmoid soft clipping function and selected the maximum as the fusion coefficients.For the high frequency subbands,the improved spatial frequency and the Laplace energy are calculated as input and link strength of the PCNN model,respectively.Then selected the maximum pulse number as the high frequency subband coefficients.The experiment results show that compared to the wavelet and NSCT image fusion algorithms,the proposed algorithm achieves better results both in subjective visual effects and objective criterion.