Solution of 2D Electromagnetic Inverse Scattering Problem Using Iterative Shrinkage-Thresholding Algorithms

SALSA targets a dual problem which achieves the same minimization as the sparse optimization problem. The dual problem splits the unkown vector t into two unknowns d and v . The first vector is defined for the data misfit minimization while the second is defined for the regularizer penalty term. The sparsity optimization problem is achieved under th constraints that these two vectors are equal. Truncation At the beginning of BIM iteration, the approximation of H 0 matrix is far from the actual H . Hence, the sparse iteration is computed only for few iterations to compute the most significant components which are assumed to be larger than the error and noise level. As the BIM iteration proceeds, H p matrix become more accurate. Consequently the number of the regularization iterations is increased. Formulation • 2D Electromagnetic Equations: