Compressed Sensing Image Processing Based On MCA Image Representation

Traditionally,information processing must meet Nyquist sampling theorem that the sampling rate is two times larger than the signal bandwidth at least.The increase of the amount of information data has brought great pressure on the traditional signal acquisition process.Compressed sensing(CS)theory proposed by D.Donoho and E.Candes in 2004 years breaks the Nyquist sampling theorem,and now it has been used in signal processing widely.The research in the paper is about the application of CS theory in image processing.(1)Curvelet-FIST algorithm based on CS theory frame is proposed and used for image super-resolution(SR)application.In the algorithm,Curvelet basis is used as sparse basis for the image,and a pseudo-star-shape sampling matrix is also proposed as measurement matrix.Fast iterative thresholding algorithm(FIST)is then employed to recover the measured signal.The simulation results show that the acquired SR image through Curvelet-FIST algorithm is better than through conventional interpolation algorithm and Wavelet-FIST for the different type of images.(2)According to the embodiment of the component learning,a complete image contains a lot of different ingredients,so it appeared error inevitably when using a single base function to sparse the image.So the image should be expressed as a combination of a number of different sparse functions.An image processing method of compressed sensing based on MCA image representation is proposed and applied to the SR image restoration and reconstruction.The acquired SR image is better than the case of a single sparse representation.On the one hand,MCA method used to represent the image can be more accurate.On the other hand,it has smaller amount of image data processed compared to the original MCA method because CS theory frame is used.