A new Success Estimation Method (SEM) for image unmixing in spatially varying single-path mixing scenarios combining attenuation and spatial distortion, is presented. Staged Sparse Component Analysis is used for estimation of the mixing model and separation of the images. SEM, relying on the assumption of sparseness, inspired by the mask reconstruction method that is used in under-determined systems, is then introduced in order to close the loop and refine the source separation. The method is compared with previous known methods, demonstrating its superiority. An optimization scheme that utilizes the SEM as the cost function is used in order to refine the system vector of parameters and improve, in turn, the final source reconstruction.