Vanishing-Point Detection based on New Perceptual Strength of the Line

Abstract

Parallel lines in the real space(3D space) are converging at the point during the projection on to the 2D space by the lens of camera, by the linear perspective. We call this point as a Vanishing-Point and this point indicates that the furthest point from the view point. Vanishing-point appears in man-made structure because it is generated by the lines, and this position can be in or outside of the scene by the scene composition. Therefore, General vanishing-point detection algorithm has to detect the point by using the only information of the image and regardless of the position of true vanishing-point But almost of the recent research for detecting vanishing-point are targeting specific scenes such as road scene, corridor scene for Auto navigating system and 3D building structure restoration those algorithms having a limitation as like vanishing point has to be located in the image, camera parameter is needed. In this paper, our research team is targeting the image which is having a vanishing-point and do not use additional information, detection regardless of the position of true vanishing-point. We are focused on line based vanishing-point detection and better measure for strength of the line. so that we suggested Pattern Matching and New Strength Measure as better measure for strength of the line. Pattern Matching is decision method whether the edge pixel is contributing to make a line or not, New Strength Measure is perceptual compensation of the strength method based on man-made structure scene which is having a periodic line. To justify the efficiency of the proposed method, We choose 100 images which is having a vanishing-point more than one. and We detect the vanishing-point by using several existing algorithms and proposed algorithm.

Key words : Vanishing-Point, Linear perspective, Pattern matching, Strength of the line, man-made structure Student number : 2012-20764