ALSD: An algorithm for noise erasure and structural information recovery for surfaces with structural discontinuities

Image inversion techniques revolving around Tikhonov regularization try to use local heuristics to clean the noise and recover the image. We propose a segmentation based inversion algorithm which can recover the structural information in a two-dimensional image data composed of continuous sub-regions with discontinuous and disjoint adjacencies. We present results for different kinds of images with different noise and structural characteristics. We demonstrate that the algorithm works with a variety of noise signatures, is parallelizable and scalable.