Image Recovery Based On Compression Perception

Image as one of the important mediums in modern information transmission system, it has become an indispensable method to capture information for people in daily life with the development of imaging technology, image processing technology and the network technology. The rapid popularization of the high-definition camera, smart phones, tablets and other consumer electronic products makes obtaining high pixel image easy. More and more image information is spread in daily life and it has brought new challenges for the image information storage, transmission and processing.The traditional Nyquist sampling theorem has dominated the digital signal processing field over the past few decades. But with the rapid growth of information transmission, the traditional Nyquist sampling theorem has already cannot satisfy the needs of the development of information. So seeking a new method that can be used to rebuild the original signal with low sampling rate becomes a kind of urgent need. The Compressive Sensing can meet the requirements of the sampling theory and technology.This paper revolves around the CS theory and introduces the principle and processing of the CS theory. Then, the paper explains the basic image restoration algorithm which was proposed in recent years and analyzes the advantages and disadvantages of the different algorithms through some experiments. That contains the OMP algorithm, the ROMP algorithm, the SAMP algorithm and other ect. At last, the paper puts forward an improved algorithm named AROMP algorithm for the drawbacks of ROMP algorithm that isn’t adaptive sparse degree. AROMP algorithm takes the advantages that the ROMP algorithm filtrate atoms two times to ensure the accuracy with the advantages of adaptive sparse degree in the improved SAMP algorithm. First. AROMP algorithm takes the method of atoms matching sequencing estimation to estimate initial sparse degree and take this estimate as the initial value. Then, it updates the sparse degree by variable step length in the iterative processes. At the same time. AROMP algorithm improves the stop condition of iterative processed through the energy difference of the residual energy. It not only improves the recovering efficiency but reduce the dependence to sparse degree.Through the experiments we can find the Peak Signal to Noise Ratio and the Relative Error of the AROMP algorithm is similar to ROMP algorithm, but recovery time become smaller than ROMP algorithm when the sparse degree becomes more. And the adaptive sparse degree avoids the accuracy drop, if we don’t know the sparse degree. The ROMP algorithm is more efficient than the SAMP algorithm that is also adaptive sparse degree. Especially under the condition of sparse degree is bigger, the recovery time is far less than the SAMP algorithm instead AROMP algorithm.