A SVD-Based 2DOMP Algorithm for Compressed Image Sensing

We use the singular value decomposition of the separable measurement matrices to obtain the optimized separable reconstruction matrices and optimize the measurements. A two-dimensional orthogonal matching pursuit optimization algorithm based on singular value decomposition is proposed. Numerical experiments demonstrate that our proposed 2DOMP-SVD algorithm can significantly improve reconstruction quality and signal to noise ratio. Moreover, separable imaging operator arise naturally in many optical implementations and can satisfy the requirements for both the measurement matrix and the reconstruction matrix individually. And this design is suitable for general separable linear system.