The Research Of Remote Sensing Image Compression Based On The Wavelet Transform Of SPIHT Algorithm

Remote sensing image is a very important information source in the geographical information system. Nowadays the remote sensing image is used increasingly not only in the field of land and resources investigation which are closely related to people's life, but also plays an important role in the aspects of military. Due to their huge magnitude of data, along with the fast development of remote sensing technology, the remote sensing image data are mounting up quickly. The limited channel capacity can not meet the demand of transmitting the great quantity of remote sensing data. The data compression technology, as an effective way of solving this problem, is becoming more and more important in the field of remote sensing. This paper has discussed the technology of remote sensing image compression by the way of the wavelet transform of SPIHT.Firstly, three kinds of general compression algorithms for remote sensing images are presented, which are transform compression, vector quantization and prediction compression algorithms. The comparison of several typical compression algorithm standards such as JPEG and JPEG2000 is given. Different of the general nature images, the space partial correlation of remote sensing image is week, which contains rich information and has a high information entropy, so it is difficult to obtain a large compressed lossless ratio. After the decomposition of remote sensing images by wavelet transform, the image energy is mainly concentrated in the low-frequency part, while the energy of the horizontal, vertical and diagonal part is fewer, that is to say the wavelet transform coefficient of remote sensing images has high local relevance. Therefore the use of wavelet transform to compress the remote sensing image will get better results. Another characteristic of remote sensing image is its complex texture. When compressing the remote sensing with complex texture using JPEG block compression method, the effect of blocks is obvious. The compression method based on wavelet transform can overcome the effects of blocks restructuring images. So, this paper focusing on the wavelet transform and the lifting scheme of the second-generation wavelet transform, specifically analysis and compare the algorithm of EZW and SPIHT which both based on wavelet transform. And doing experiments to make simulation is given in the paper.This paper focuses on SPIHT algorithm, which is to achieve compression by building zero-tree, similar to the principles of the EZW algorithm using progressive transmission coding theory. The progressive transmission coding theory is arraying the absolute of value from large to little, and transmitting the important value first, which can make the quality of reconstruction images get better gradually. SPIHT organizes the wavelet coefficients together in the form of zero-tree, and suppose that if the value of the parent node less than quantitative numerical threshold T, the probability that all the coefficients of tree nodes less than T is large. With only one bit can represent the value of the current coding plane, we can achieve the compression effect. Although the SPIHT algorithm achieves good compression efficiency, there are still many points need to improve. First, SPIHT algorithm based on the floating-point operations of Mallat decomposition which requires a large amount of memory modules, and it is a great obstacle for hardware to realize high-speed. Second, it needs three lists (LSP,LIP,LIS) to preserve some information of splitting during both encoding and decoding which is hard for hardware implementation. Third, the algorithm is only using the similarity of the same direction of the sub-bands, but overlooking the correlation of the internal coefficients.Taking the shortages of the SPIHT into account, the method that combining the second generation wavelet transform coding and SPIHT is proposed. Because lifting wavelet is more effective in focusing the image energy than traditional wavelet, using it can improve the reconstructed image quality. Using MATLAB and VC++6.0 for simulation, by biorthogonal 9/7 wavelet, analyzing the experimental results can find that the programme can achieve good image compression effect mostly, but in the case of low bit rate, especially for high-resolution remote sensing images, the large number of details are lost easily. The effect of the reconstruction image is not satisfactory.To solve this problem, the improved algorithm is proposed on the base of programme one framework. It is based on SPIHT, and improves the structure of zero- tree and the procedure of coding. In the meanwhile, it broaches the mind of LZC algorithm's symbol flag. The improved algorithm uses more effective zero-tree, which not only makes full use of the correlation of the decomposition of wavelet decomposition between sub-bands, but also uses the relevance of adjacent coefficients in the low frequency band. It overcomes the shortages of coding all coefficients in the low-frequency sub-bands independently at each quantization threshold, which is increasing coding efficiency. In addition, the improved method adopts the strategy of code scanning and refinement code crossing, It makes the refinement coding of important wavelet coefficients embedded in the coding process, which enables more and more bits code for important wavelet coefficients, enhances the quality of reconstruction image in low-bit. Simulation results show that the improved programme in the condition of reconstructing the quality of images, reduces the memory consumption, increases coding rates. Compared with SPIHT algorithm, the improved method has excellent performances such as the peak signal- to- noise ratio when reconstructing images and the quality of visual images, which is more obvious in the low bit rate case.