Satellite Remote Sensing Image Fusion Based On Compressed Sensing

In recent years, along with the rapid development of space technology, remote sensing satellites that carrying a variety of sensors have been launched, which makes it extremely wide range of different spatial resolution satellite remote sensing data. Especially, the China’s Beidou navigation satellite network success, which means the research to how to get more richer, useful and reliable information from these remote sensing images is become the focus of the current field of Remote Sensing Applications need to be addressed urgently. Multi-source remote sensing image fusion is defined as the technology to process the date of different sensors in the same area in order to improve the quality of access to information and achieve information superiority complementary. As an important part of the project "xx ", this article study the high-resolution panchromatic image and low-resolution multi-spectral image fusion. Base on the classic fusion algorithms on remote sensing image fusion, the IHS transform, PCA transform and discrete wavelet transform, we carried out in-depth research. The Compressed Sensing was first proposed in2006and unprecedented rapid development recent years shift the sampling of the signal to sampling information abandoned the full sample greatly reducing the potential consumption of traditional signal acquisition and processing flow. We innovated combined compressed sensing and satellite remote sensing fusion algorithm and proposed a new fusion algorithm. The main work done as follows:The paper studies the registration technology before the satellite remote sensing image fusion firstly. And proposed registration algorithm based on SURF theory. It is a new fast interest point detection and representation method. The algorithm introduced in the integral image and template approximation, and detect feature points by flash Hessian detection sub. First-order Haar wavelet response is used to determine main direction and calculated64dimensional feature point descriptor in the characterization stage. It match feature points between images according to the Euclidean distance. Simulation results indicate that this registration algorithm is better than the traditional registration algorithm to ensure accurate registration while the real-time performance.Then we discuss the satellite remote sensing image fusion algorithm based on compressed sensing theory. Introducing compressed sensing theory, we proposed two fusion algorithms:the CS-IHS algorithm and the CS-FWT-PCA algorithm based on traditional IHS transform, PCA transform and discrete wavelet transform fusion method. We use "Daubechies13" wavelet as sparse base, and part of the Hama Da Matrix as measurement matrix, and SAMP reconstruction algorithm as the reconstruction algorithm in the CS-IHS algorithms, and coefficient absolute value of the largest law as the fusion rules. In the CS-FWT-PCA algorithm, we introduced of symmetric fractional B-spline wavelet as sparse base, and still use Hama Da matrix as the measurement matrix and SAMP reconstruction algorithm as the reconstruction algorithm, using the improved and based on the local variance as the fusion rules. The experimental simulation results show that the two algorithms achieve better fusion effect than traditional fusion method, which the fusion effect of the fusion algorithm based on CS-FWT-PCA is better.