Research On Remote Sensing Image Fast Compressing

With the emergence of big data problems, in which the size of the processed data sets can easily exceed terabytes, the "fast" in FFT is often no longer fast enough. Sparse signal processing has recently witnessed two major research trends, namely compressed sensing and Sparse Fast Fourier Transform(SFFT) at home and abroad, which were on the list of top 10 high technologies that could probably change the future life in the "technical review" of the United States in 2007 and 2012, respectively.Based on the theories of compression sensing and SFFT, this article studies the remote sensing satellite images. Three classical reconstruction algorithm of compression sensing, OMP, SAMP, and StOMP, were carried out in-depth. Besides, we try to put forward an improved algorithm, and innovatively apply the SFFT for the remote sensing image compression, in order to accelerate the reconstruction process while maintaining high image quality. The main work done as follows:First we compare the performance of three classical reconstruction algorithms comprehensively of compression sensing, that is OMP, SAMP and StOMP. Taking both efficiency and quality into consideration, this paper try to improve the performance of StOMP algorithm, and innovatively put forward a novel algorithm StOGP, which combined StOMP with Gradient Pursuit. The algorithm take advantage of adaptability of the StOMP algorithm, whose speed is higher and even with unknown sparsity of the K, it can gradually get close to the input variables by the step strategy and just choose a few atoms to update the support set each time. Gradient Pursuit, which can obtain the optimal solution at the direction of the negative gradient, was introduced to update the rest variables. Experimental simulations show the proposed algorithm performance better than traditional algorithm both in image quality and the real-time capability.Then we discuss the theory and the process of implementing of SFFT algorithms in detail, and measure the performance of three existed SFFT algorithms through the simulation experiment. Then apply the improved SFFT3 algorithm, which contains the best comprehensive properties, for the remote sensing images compression. Compared with compression sensing based on the reconstruction algorithm StOMP, whose reconstruction time is the least, and SAMP algorithm, whose reconstruction images quality is the best along with the three classical reconstruction algorithms discussed before, the results show that reconstruction time of image based on the improved SFFT3 algorithm was far less than the CS based on StOMP algorithm, and at the same time, the reconstructed image quality is higher than the CS based on SAMP algorithm. Therefore, it can be concluded that the SFFT algorithm has a good prospect of signal processing in the future.