期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2017
卷号：95
期号：1
出版社：Journal of Theoretical and Applied
摘要：Orientation map is an image representation method representing edges of objects in an image as a set of short line segments instead of pixels representation. The orientation map approach not only aims to extract the key orientation information which constitutes the semantic framework of the visual scene, but also to remove the insignificant elements which barely play a role in image understanding. However, there are two main problems in the current computational approach of orientation map: 1-the outliers in the orientation map 2- the deformities in some line segments. These two problems have an adverse impact on any algorithmic work for image interpretation. This article proposes two improvements, heuristic based and genetic based, to reduce outliers and deformities respectively. Experimental results prove the significant improvement on the current computational approach of calculating the orientation map.
