An automatical algorithm based on local image properties

摘要

This study introduces the definition of local image properties. Using so-called local rules one can describe recognitions of interesting foreground object characteristics in images. We propose a novel algorithm to automatically detect interesting object characteristics in the images with local image properties, including the detections of edges or corners of objects in images. As applications, firstly, we compare the detection results of edges for objects in four images by our algorithm with the Roberts operator, Sobel operator, Prewitt operator, Log operator, Canny operator and cellular-neural-network-based algorithm (CNNBA). Secondly, we compare the detection results of corners for objects in three images by our algorithm with Harris operator, Shi-Tomasi operator and CNNBA. The experimental results have demonstrated that the proposed algorithm is able to detect correctly the all edges or corners of the objects in corresponding images, and has superior performance when compared with the other methods.