Hyperspectral Image Compressed Sensing Model Based On The Collaborative Sparsity Of The Intra-frame And Inter-band

Hyperspectral remote sensing as a new technique developed in the 1980 s,with the development of imaging spectrum technology,it canrecord more and more narrow continuous spectrum data in the wide range of electromagnetic wave.These abundant data information makes the hyperspectral remote sensing image used in agriculture,forestry,atmosphere and environment monitoring and other fields and play an important role.But the obtained surge data of imaging spectrometer from the space-borne equipment,it brings huge burden for data storage and real-time transmission,therefore,the research of hyperspectral remote sensing image compression technology become more important in the field of remote sensing.The traditional compression process is the Nyquist sampling of hyperspectral image,and then carry on some transformationto obtain a large amplitude coefficientand and to give up the lower value coefficients,this sampling and compression method can not meet the needs of space-borne equipments,however,the compressed sensing theory can solves this problem well,guided by the CS theory and the optimization method,it realizes the sampling as well ascompression,and can obtain a few observationsin the encoding side,then can use these observations to carry on the high probability mage reconstruction for hyperspectral in the decoding side.This paper has done research on the compressed sensing model of the hyperspectral remote image,innovation pointsperformance in the following two aspects:Firstly,in the framework of the theory of compressed sensing,the Gaussian random matrixwas selected as theobserved matix to carry on sampling andcompressing for the hyperspectral remote sensing data in the same time to obtain the fewer sampling values,and then transmitthe sampling values to the reconstruct image decoding device on the ground through the channel.Based on the RCoSrecoveryalgorithm which considers the local smoothing sparse and nonlocal self-similarityin the natural images,this paper proposes a new model,which is the hyperspectral image compressed sensing model based on the collaborative sparsity of the intra-frame and inter-bandframe.This model considering the three aspects of sparsity of the hyperspectral remote sensing image as a priori knowledge to constraint the objective function,the three sparsity included the local smoothing similarity of hyperspectral image,the non-local self-similarity and the similarity between thespectrum bands.Through a large number of experiments,the reconstructed image quality is significantly higher than the existing GPSR algorithm,the improved algorithm GPSR algorithmand RCoS reconstruction results using the propsed model HICoSM's recovery algorithmin this paper.The second part is based on the first part and it is the improvement of the first part.The model considers the local smoothing similarity,the local self-similarity and spectral correlation atthe same time.In the optimization process,it uses 3D-TV model to solve the local smoothing sparse and the spectral similarity together,and then combined with the prior condition of the nonlocal self-similarity,and finally uses the separation Bregman iterative algorithm to find the the reconstruction results in the optimizationprocess.This paper uses 3D-TV model to improveHICoSM,aiming to recove the hyperspectral remote sensing image effectively and at the same time to have the repair and the denoisingeffect.Through the experiment provesthe effectiveness of the jointed model based on HICoSMcollaborative sparse model and 3D-TV model.Compared to the classic sparse gradient projection algorithm,the sparse gradient projection algorithm based on linear filter between the spectrum bands,and the RCoS algorithm,it shows that theproposed algorithm have ancertain degree effect on improving the restorationquality of the hyperspectra image.Though mining and analysis for a series of the prior knowledge of the hyperspectal image,this paper proposes a new model,and uses the 3D-TV to improve it,and achieve the better recongstructed results compared with the some present classic algorithms,the experiments prove that this algorithm have an effect on improving the hyperspectral image reconstruction quality.