Biomimetic Design for Efficient Robotic Performance in Dynamic Aquatic Environments - Survey

This manuscript is a review over the published articles on edge detection. At first, it provides theoretical background, and then reviews wide range of methods of edge detection in different categorizes. The review also studies the relationship between categories, and presents evaluations regarding to their application, performance, and implementation. It was stated that the edge detection methods structurally are a combination of image smoothing and image differentiation plus a post-processing for edge labelling. The image smoothing involves filters that reduce the noise, regularize the numerical computation, and provide a parametric representation of the image that works as a mathematical microscope to analyze it in different scales and increase the accuracy and reliability of edge detection. The image differentiation provides information of intensity transition in the image that is necessary to represent the position and strength of the edges and their orientation. The edge labelling calls for post-processing to suppress the false edges, link the dispread ones, and produce a uniform contour of objects.