Compressed Sensing Algorithm For Image And Its Application Based On Variable Sampling Rates And Prediction

Compressed Sensing(CS) is a new method that can be used for the sampling and reconstruction of a signal. It breaks through the traditional Nyquist sampling theorem, and can use fewer samples to represent the original signal, then uses these samples to recover the original signal accurately. CS can complete the sample and compression of a single at the same time, but it requires the single to be processed sparse. Compared with the traditional signal processing methods, CS theory to some extent improves the sampling efficiency of a signal, and reduces the number of sampled values required to recover the signal. However, if CS obtains the measured values of an image signal still by a unified sampling rate and random sampling, its sampling efficiency can't reach a deal result. Therefore, this paper improves the existing CS method, and combines with the human visual characteristics, the contribution quantity of different image regions to the recovery accuracy and the prediction algorithm to propose a novel CS scheme adapted to the image signal.The main works and innovations are outlined as follows:Firstly, in order to solve the problem that the sampling efficiency of CS is low in processing an image signal, a novel CS scheme for image based on variable sampling rates and prediction is proposed. Combining the visual characteristics of human eyes, this scheme uses different methods to deal with the different parts of the image, that is, image edge regions are allocated more samples, and non-edge regions extracted in accordance with certain rules are allocated fewer samples, and then the non-edge regions un-extracted is obtained by prediction. Simulation results show that the proposed scheme can have an improvement of 1.4d B in recovery quality compared with existing algorithms when the amount of samples is same, the image structure similarity is higher, the texture details of the image are clearer. Therefore, samples the proposed scheme required are less, and its sampling efficiency is higher. In addition, in the same software and hardware environments, the processing time of the proposed scheme can be decreased by at least 50% compared with the existing algorithms.Secondly, in view of the problem that the information quantity of spectral image is very big,the satellite link resources are limited, and the correlation between adjacent spectral bands of spectral images, this paper utilizes difference model and the novel compressed sensing scheme this paper proposed to deal with spectral images. Only a basic spectral band and some difference spectrum bands needed are to deal with, and the sparse degree of the difference spectrum band is better, thus it needs fewer samples to recover the original image. Simulation results show that the proposed CS algorithm based on the difference model can have an improvement of 2.1d B in recovery quality compared with the other two methods. Therefore, the novel CS algorithm based on difference model can effectively compress and transmit the spectral images in the inter satellite links.