Compensation of the effects caused by the outliers in multiframe iterative techniques

In this work the compensation to a problem in the iterative multiframe regularized least-squares (RLS) technique is proposed. The use of a sequence of frames presenting outliers is possible, as long they are properly estimated and removed. The removal of the outliers causes an unequal contribution from the additional images in the estimation. The unequal contribution complicates the search for a solution in the gradient projection iterative techniques. As rescaling eigenvalues requires a high computational cost, the compensation of the spatial contributions becomes an effective solution. A compensation matrix is used in the iterative technique. The compensation is based on the inversion of the effects of the degradations in the image components of constant amplitude. Moreover, good results are obtained soon at the first iterations, allowing earlier termination with satisfactory results.