RE-SAMPLING BASED IMAGE CODING FOR REMOTE SENSING IMAGES

摘要

The resolution and data rate are getting higher and higher with the development of remote sensing technique. In order to solve the contradiction between the mass data rate and the available limited channel bandwidth, real time on-board image compression technique is necessary for remote sensing images. Several special features, high fidelity of restored image (means visual loss-less) and low complexity of compressing algorithm (means easy to make a on-board real time compression system with compact in structure, low weight, low power consumption and higher reliability), are required for on-board remote sensing image compression technique. Wavelet based image compression method is a generally accepted best image compression method. But for the case of on-board real time compression, the method has obvious weakness. Because in this case for getting better compression results large processing unit and multi-stage floating coefficient wavelet decomposition are required and it is difficult to implement in compact hardware. Since wavelet based image compression methods have above mentioned inherent weakness and it is difficult to make a compact and reliable real time compression system for on-board usage, a new image compression method is presented in this paper. It is important to notice that to human visual sensing, gray level is more important than spatial resolution in soothing areas and it is opposite in texture areas. Making better use of the above mentioned human visual sensing characters, a new image compression method named Re-sampling Based Compression (RBC) is created which is characterized with no image data transformation, having only integer operations (no floating point operations) and only small processing unit have to be used. The RBC is not only succinct in compressing algorithm but also high fidelity for restored images, so it is quite suitable for high speed on board remote sensing image compression. A comparison of the compressing results between RBC and SPIHT (Set Partitioning In Hierarchical Tree) is given in this paper. The results show that, within the range of vision loss-less (compression ratio less than 10) and under the condition of about equal quality of restored images, not only the required processing unit of RBC is much smaller than that of SPIHT but also processing speed of RBC is much faster than that of SPIHT and the RBC is more suitable than SPIHT for remote sensing usage is proven by the results.