Food image representation based on modified orientation map
Salwa Khalid Abdulateef, Massudi Mahmuddin, Nor Hazlyna Harun · Universiti Utara Malaysia Institutional Repository (Universiti Utara Malaysia) · 2017
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.