Research And Implementation Of Image Compressive Sensing Based On Joint Priors

The Internet of Everything is a necessary part of building smart cities,smart manufacturing and other scenarios,and its demand for high-definition,high-precision data in many forms puts tremendous pressure on storage and transmission.Compression-awareness compresses data while sampling,which can effectively reduce the required sample size to provide a new direction and solution for the Io T to solve the acquisition and transmission problems.In this paper,an in-depth study of image compression perception is done for three major elements: measurement matrix,sparse dictionary,and reconstruction algorithm,and the main work includes:(1)The working theory,the underlying mathematical model and the key techniques in compressive sensing are described.Existing optimization methods for measurement matrices and sparse dictionaries as well as reconstruction algorithms are analyzed,and the problems and shortcomings in existing research results are studied and analyzed.(2)For the optimization of measurement matrix and sparse dictionary,a modified alternating optimization method of measurement matrix and sparse dictionary is proposed considering that the measurement matrix and sparse dictionary are coupled and affect each other.First,the sparse dictionary is fixed,and the adaptive gradient descent method is used to make the Gram matrix infinitely approximate the unit matrix and the matrix determined by the structure of the sparse dictionary to obtain the optimized measurement matrix;then,the sample data obtained from this matrix is used for sparse dictionary learning,and the result is used as the dictionary input for the next round of measurement matrix;finally,the peak value is improved by iterative update without increasing the computation time.Signal-to-noise ratio level.(3)Combining the sparse prior and the image statistical prior information obtained from natural images,a joint prior information model is designed and solved using an alternating solution method to solve the joint prior information model.Firstly,we fix the reconstruction result and use the noise energy as the threshold to solve the local sparsity subproblem using the OMP algorithm to obtain the sparse coefficient matrix;then we fix the sparse coefficient matrix and introduce the semi-quadratic splitting method to solve the statistical prior information subproblem with the effect of local sparsity to obtain the reconstruction result;finally,we iteratively update it to obtain a better reconstruction result and less affected by noise.In this paper,the effectiveness of the alternating optimization method of measurement matrix and sparse dictionary,joint prior information reconstruction model and joint prior information reconstruction algorithm proposed in this paper are verified by simulation experiments with real images as test samples,and the performance of the compressed sensing system can be improved under the same conditions compared with the experimental comparison algorithm.