An Amalgam Approach to Detect Edges Using Ultrametric Contour Map in Natural Scene Images

摘要

Detection of straight lines in digital images is a reappearing problem in image processing and computer vision tasks. Typically the edges or lines in an image are considered as a very influential component which carries crucial information about the whole image. This important component can be extracted by employing various edge detection methods. Generally, the lines are formed by boundary pixels which connect two isolated regions, hence it easy to detect such edges between the regions. Still, it has some obstacles to detect weak edge pixels that form a continuous straight line in natural scenes. Here in this paper, we proposed a method to detect straight lines in such natural scene images which contain “Roads”, “Bridge”, “Pavement”, and similar type of archaeological and natural structures. Such structures in photographs start from the bottom of the frame and move towards some region, object or meet at a vanishing point. The proposed method has been tested on our own dataset of natural scene images and we have seen that our proposed method to detect lines for some specific images performs much better upon current state-of-the-art in the line detection techniques.