Image Analysis by Structural Dissimilarity Estimation

Image analysis using dissimilarity estimation methods is a crucial process in several image processing applications. For instance, it is used in frame reconstruction, pattern recognition and image inpainting. Dissimilarity estimation methods based on intensity information lead to impressive results with different image types, but failure cases still exist especially when dealing with local structural variations. In this paper, a metric for dissimilarity estimation, based on structural information, is proposed. Results show that using the proposed metric, local dissimilarities can be better estimated comparing to traditional intensity-based metrics, thus helping the analysis of images in remote sensing images.