Edge map improvement on Spiral Architecture

Edge map is considered as an important entity containing most of object features in an image. Many computer vision systems rely on the use of the boundary line information to perform the object recognition tasks. However, with the exception of images acquired from highly restricted environment, common edge detectors do not guarantee the production of continuous boundaries of objects. In this paper, a local processing based edge-linking algorithm is proposed. We set a criterion involving multiple properties of the pixel to find the best candidate points for edge linking. In addition, this algorithm is implemented on Spiral Architecture