Research On Image Reconstruction Based On Compression Perception

With the age of information coming,the image signal plays a vital role in the domain of information transmission,and the people's requirements for images are constantly improving.Compressed sensing(CS)theory,as a new signal sampling processing technology,compressing the signal to its sparse domain by sparse transformation,and then realizes efficiently signal acquisition and recovery processing.It is widely used in the field of signal and image processing.Based on the theory of CS,a fast and accurate image sparse reconstruction algorithm is designed and applied to the synthetic aperture radiometer and MRI image's reconstruction,and has achieved good image reconstruction results.The thesis' main research contents are as follows:(1)The SL0 inverted optimization algorithm is optimized for the SAIR,the original Gaussian smoothing function is improved to a composite function with better approximation effect.A fast CS-L0 algorithm based on the two-dimensional reconstruction model is proposed and applied to the image reconstruction of the synthetic aperture radiometer(SAIR),which get a good reconstruction effect.On the basis of the SAIR two-dimensional dual D imaging model,the SAIR image is reconstructed by the advantage of two-dimensional model and smooth zero norm solution.The simulation shows that the CS-L0 method based on two-dimensional model effectively improves the PSNR and SSIM index of the reconstructed image of SAIR millimeter wave than traditional CS inverted algorithm.The PSNR of reconstructed image is increased by about 0.6 d B,the SSIM of reconstructed image is increased by about 0.05,and the inversion speed is improved by about 20%.(2)To take advantages of Shearlet transformation(ST)multi-scale decomposition and TVAL3 algorithm for MRI image reconstruction,a CS reconstruction algorithm based on discrete shearlet transformation is designed-the STTV algorithm,and has achieved a good reconstruction effect in the image reconstruction of magnetic resonance imaging(MRI).The STTV algorithm uses discrete shearlet transform to decompose the MRI image in multi-scale sparse,then reconstructing the high-frequency subband information by TVAL3 algorithm.Finally,the reconstructed high-frequency and original low-frequency information are inverted to obtain high-quality MRI reconstructed images.Compare with the STOMP and wavelet TVAL3 algorithm,the proposed STTV algorithm has higher reconstruction accuracy,the reconstructed MRI image is more accurate and clear.The reconstruction of STTV algorithm MRI the image PSNR is raised by about 1 d B,and SSIM's improvement is about 0.05.In addition,the PSNR of reconstructed MRI image with noisyis about 1.5 d B higher than the DWTTVAL3 algorithm.