Research On Compression's Technology Of Hyper-spectrum Image Base On Lift-wavelet And Implement On DSPs

Remotely sensed hyperspectral image is a 3D stereoscopic image, that is to say, having another dimensional spectrum information again on the foundation of common and two-dimensional picture. So the data of remotely sensed hyperspectral image is huge and it is hard to deliver and saving directly. So compression to the image is necessary. Because the remotely sensed hyperspectral image information is very precious, the lossless compression or near have lossless compression method is needed possibly. So far, the super spectrum picture hasn't been becoming a set of mature or standard compress technique. Therefore the research of compress coding to the remotely sensed hyperspectral image has the important applied value.Aimed at currently the remotely sensed hyperspectral image two-dimensional picture compress an existent problem the thesis put forward a kind of super spectrum picture compress project without the lossless. The project considers the special of the remotely sensed hyperspectral image and the realization of hardware, so the remotely sensed hyperspectral image two-dimensional picture compress will adopt the method based on LeGall5/3 wavelet transform. There are some advantages of this method such as: simple moves and addition operation, quick speed, occupy a little memory and suitable for removing the remotely sensed hyperspectral image data redundancy than the generally small wave transformation. It is specially suited for the situation which is real-time, high speed code or losses compression and benefits the hardware realization.Picture after wavelet transform coding, Embedded Zerotree Wavelet (EZW) build up frequency band to take wavelet transform sparse relation to reduce the redundancy information of the wavelet transform coefficient at current threshold, thus in new way the organization become a data to flow. Pass again Arithmetic Coder to complete end coding, flow data to compress to flow for the bandwidth compression data. The EZW coding can stop when the whichever threshold, and at present once coded stiller than flow especially can rebuild to correspond a picture of protecting the true degree, along with decoding the code flow in the gradual increment of the quality layer, the solution presses out of picture PSNR to also enlarge gradually, the picture is more and more clear, keeping to decode the picture lossless This provided a good compression solution for the hyper spectral image. When picture quality's request for us low, can spread to compress code to flow a medium lower layer by, being to have higher request to the picture quality, can continue to deliver a higher layer, carrying out real meaning of press need to be delivered, save a precious network resources effectively, in the meantime keep all original information. Experiment figure out that applying LeGall5/3 wavelet transform and binary arithmetic coding to combine under the equal condition, compress rate for about 2, surpass that of JPEG-JS 10% at least.The method be realized with wavelet transformation on DSP chip(TMS320DM642), the experiment use CCS and the TMS320DM642 EVM board, to optimize the parts of EZW, processing a size of 512*512 of the hyper spectral image needs about 11 seconds for lossless compression, decrease about 40 percent before optimizing, it proves that this method can be realized on the DSP chip.