Research On Image Processing Based On Compressed Sensing

Firstly, some basic idea about Compressed Sensing (CS) theory is discussed, which includes measurement matrix, sparse representation of transform space, and common image reconstruction algorithms. Secondly, greedy algorithm and convex optimization algorithm are introduced and simulated to approve the effectiveness and superiority of OMP algorithm. Lastly, two aspects of application by CS are studied, which consists of image encryption and feature detection.In the image encryption, sparsity and reconstruction, which play a huge role in CS, are added to the chaotic encryption. The detailed procedure is to initialize the image by wavelet transform, which leads to a sparsified representation and measurements. Then the chaotic method for linear measurement is applied to encryption and decryption. At last, the encrypted image will be reconstructed by the inverse wavelet transform. This method, other than increase the time of encryption and decryption of the image, reinforces the difficulty of being decrypted, thusly improving the security of the encrypted image. Image feature detection is refers to the use of detection of parametric shape by CS. The detection of line is detailed in the paper by Hough transform, which utilize the parametric model of line and map every pixel of image to parametric space, and then find the peak in the space. That’s to convert the problem of finding the spatial distribution pattern into detection of peak in parametric space.