Astronomical Image Compression And Denoising Reconstruction Algorithms Based On Compressed Sensing

With the development of aerospace science and technology,to expand the living space and search for extraterrestrial energy become increasingly pressing for human being.Deep space exploration has become the focus of the current and future development of the word.Astronomical image plays an imp ortant role in deep space exploration since which can directly reflect the astronomical information.The important information,such as the geographical environment?whether there is extraterrestrial life and water and so on,can be directly obtained from these received astronomical images.However,due to the received astronomical images are usually high resolution images,traditional compression methods are difficul t to obtain a high astronomical image compression ratio,resulting in longer image transmission time.In addition,astronomical image is often disturbed by cosmic noise in the process of image transmission,it is difficult to analyze these received astronomical images as these images received by the ground station usually contain a lot of noise.Compressed sensing theory can achieve signal compression in the process of signal sampling,not only that,it can reconstruct original signal with high quality using only a small amount of compressed data.The CS theory is applied to astronomical image compression and denoising reconstruction based on its advantages in this dissertation,the major research results are as the following:The implementation process and compression principle of CS theory are introduced,and the three main parts in CS theory: signal sparse representation ?measurement matrix and the reconstruction algorithm are analyzed and studied in detail.To aim at the characteristic of astronomical image,the mathematical model is established for designing CS reconstruction algorithms in the following chapters.The compression performance of the classical JPEG and JPEG2000 astronomical image compression method are analyzed,the CS theory is applied to astronomical image compression for further improving astronomical image compression efficiency,meanwhile CS compression reconstruction algorithms are studied.To improve the reconstruction performance of iterative hard thresholding algorithm,improved block sparse total variation method with astronomical image characteristics is applied to adjust the reconstructed image in each iteration,then modified IHT algorithm is presented.The experiment of astronomical image based on CS compression reconstruction algorithm: IHT algorithm?MIHT algorithm and JPEG with human vision system algorithm and improved JPEG2000 algorithm shows that CS compression reconstruction algorithm can obtain better astronomical image compression performance in a relatively shorter time,the advantage of CS theory in high resolution astronomical image compression is fully demonstrated.The CS astronomical image denoising reconstruction algorithms based on wavelet transform are studied.To solve the problem of poor denoising effect and slower convergence speed in IHT algorithm,the astronomical image denoising reconstruction algorithm based on wavelet transform is proposed in this dissertation.The algorithm proposed firstly uses the decreasing Visual Shrink thresholding proposed to select the astronomical image wavelet coefficients,then cycle spinning method proposed is adopted to adjust the reconstructed astronomical image for suppressing the pseudo-gibbs phenomenon caused by wavelet transform without translation cycle characteristics in each iteration,meanwhile the proposed Dai-Yuan stepsize based on CS is used to adjust the algorithm convergence speed.The experimental results demonstrate that the proposed algorithm can obtain better low noise astronomical image denoising performance and faster convergence speed.The CS astronomical image denoising reconstruction algorithm based on curvelet transform is studied.Wavelet transform does not have the multi-direction property,which makes it difficult to provide the optimal sparse representation for high dimensional astronomical images with line or surface singularity.At present,the popular multi-scale geometric analysis methods can provide the optimal sparse representation for image,therefore which are used to design CS astronomical image denoising reconstruction algorithm in this dissertation.Curvelet transform is a non-adaptive multi-scale geometric analysis method,which has better sparse performance than wavelet.To improve the denoising performance of CS iterative shrink-thresholding algorithm based on TV and curvelet,an astronomical image denoising reconstruction algorithm based on curvelet transform is proposed.This algorithm proposed uses the curvelet wiener filtering operator designed to replace the threshold operator in ICT-TV method for selecting the curvelet coefficients of astronomical image,and the curvelet TV method that proposed is applied to adjust the reconstructed image for further improving the reconstructed quality of astronomical image.The experimental results show that the denoising and reconstruciton ability of the proposed algorithm has a performance improvement,which can reconstruct a high quality image from the high noise astronomical image.The CS astronomical image denoising reconstruction algorithm based on non-subsampled contourlet transform is studied.Although curvelet transform has better image sparse representation ability,it also does not have the property of cycle translation,the pesudo-gibbs will appear in the process of image denoising using the threhsolding method.The contourlet transform in non adaptive multi-scale geometric analysis method not only solve the problem of high redundancy in curvelet transform effectively,but also which has excellent sparse representation capability of curvelet transform.Additionly,NSCT has the cycle translation characteristics that wavelet transform ? curvelet transform and the traditional controulet transform lacked.When the threshold method is used for image denoising,the pseudo-gibbs effect will not appear.An astronomical image denoising reconstruction algorithm based NSCT is proposed to further improve the astronomical image denoising ability of iterative soft thresholding algorithm based on NSCT.The algorithm proposed uses the improved Baye Shrink to select the NSCT coefficients,meanwhile the iteration stop condition is modified to improve the reconstructed image quality.The experiment result shows that the algorithm proposed improves the peak signal to noise ratio in the reconstructed astronomical image.When sampling ratio is lower,the algorithm proposed can obtain better high resolution astronomical image denoising ability with less reconstruction time.