Research Of Remote Sensing Image Fusion Method

According to the problems of difficult space matching, high degree spectral distortion, limited enhance spatial resolution in the traditional pixel fusion methods, In this thesis, we put forward a remote sensing image fusion method based on the compressed sensing theory. The main research achievements are as follows:1) This paper presents a new approach to remote sensing image fusion process based on compressed sensing theory. The process combined with compressed sensing theory and wavelet transform. The multi spectral images and panchromatic images are no need to be matched.2) In this paper, we used the joint dictionary as a sparse matrix, the polynomial deterministic matrix as a measurement matrix, and reconstructed the signals with Basic Pursuit and wavelet transform method. We trained the high-resolution image patches and low-resolution image patches to obtain the two kinds of the training dictionary, which not only contains the original images structural characteristics, but also has a corresponding relationship. By this way, the restored image patches are more like the original images in some details. The polynomial deter ministic matrix can reduce the uncertainties. Basic Pursuit can reconstruct more accurate information, and wavelet transform method can separate the information, and reorganize information again. Finally, we get a reconstructed image contains more abundant information.3) We use the compressed sensing theory to get a fusion image with soft brightness, normal tones, clear boundaries, and wealth details. Therefore, the quality of the reconstructed image is higher than the images in the pixel- level image fusion methods, which has the phenomena of spectral distortion and low-resolution.