Using local structure for the reliable removal of noise from the output of the LoG edge detector

This paper tackles an important problem in image processing; that is, the detection of edges in natural scenes. A scheme that combines simplicity with the ability to detect intensity jumps at widely varying contrasts is proposed. The scheme is constructed by combining the Laplacian-of-Gaussian (LoG) edge detector with a noise removal mechanism. The mechanism is built around a proposed definition for potentially valid edge contours that incorporates their local structure in the filtering process. Some of the advantages of the proposed approach include accurate localization of the edges and ease of implementation. Simulation results as well as statistical analysis of the approach for the 1-D case are provided.<
>