Researches On Bandelets Based Remote Sensing Image Compression

With rapid development of remote sensing technology,the remote sensing images are widely used in the field of land resources investigation. The limited channel capacity can not meet the demand of transmitting the great amount of remote sensing data. Multiscale geometry analysis is a new tool that is superior to wavelet. Because of the characteristic of anisotropy and multi-directional, multiscale geometry analysis tools can represent high-dimensional singularities more effectively. Bandelets are the typical adaptive multiscale geometry analysis tools. They can take advantages of intrinsic geometric regularity of image structures and obtain the optimal representation of underlying functions. But the realization of first generation bandelets is very complicated. The second generation bandelets can avoid the adaptive quadtree partition the exhaustive searching for the optimal geometric flows so enables a fast and efficient compression.This paper focuses on the Bandelets application on remote sensing image compression. Aim at two kinds of remote sensing images, SAR images and hyperspectral remote sensing images, three algorithms for remote sensing image compression based on Bandelets are proposed. Summary the main work as follows.(1) In order to make a good use of information in each high frequency sub-band to preserve the detailed and texture information, we proposed a method based on Wavelet Packet Bandelets for SAR images. The core of this method still uses the fixed partition size, but adopt Wavelet Packet decomposition for multiresolution analysis before performing Bandelets. The simulation results show that the superiority of our method to the second generation Bandelets and Bandelets with the fixed partition size in performance, with a higher quality and a better visual effect of the reconstructed image. This is an effective method for SAR image compression.(2) Based on characteristics of sparsity of Bandelets coefficients, and apply the multiscale Bandelets with adaptive quadtree partition to encode them. Because Bandelet scan provide an efficient representation of geometric structures existed in images, the Bandelets based method can capture abundant textures in SAR images. And the coefficients decomposed by Bandelets have a similar structure with that of wavelet coefficients. Embedded Block Coding with Optimal Truncation (EBCOT) coding algorithm is employed in our proposed method. Experimental results show that our method outperforms JPEG2000 and the second generation Bandelet (2G-Bandelet) compression methods on SAR image encoding at low bit rates.(3) Some wavelet-based compression algorithms have been successfully used for some hyperspectral space missions. This paper focuses on the compression algorithm based on Bandelets. In order to preserve the detailed information as much as possible, a hierarchical compression method with protection strategy for hyperspectral images is proposed in this paper. This method is suitable for hyperspectral images compression. This is an innovating work that firstly applied Bandelets to compress hyperspectral images.