Title :
A Novel Approach for Object Detection from Scrambled Background
Author_Institution :
Sch. of Inf. Sci. & Eng., Central South Univ., Changsha, China
Abstract :
A novel approach is introduced to detect object from scrambled background. Firstly, Point-pair Set is constructed by filtrating points with similar curvature; then similar Segment-pair Set is established through searching similar segment skeleton in Point-pair Set; finally, optimal transformation is selected from the Segment-pair Set using scoring function to determine the optimal matching. Experiments indicate an encouraging detection efficiency, speed and running time complexity to irregular shapes.
Keywords :
image matching; image segmentation; object detection; filtrating points; irregular shapes; object detection; optimal matching; optimal transformation; point-pair set; scrambled background; segment-pair set; similar segment skeleton; Algorithm design and analysis; Geometry; Image reconstruction; Object detection; Pixel; Shape; Skeleton;
Conference_Titel :
Pattern Recognition (CCPR), 2010 Chinese Conference on
Conference_Location :
Chongqing
Print_ISBN :
978-1-4244-7209-3
Electronic_ISBN :
978-1-4244-7210-9
DOI :
10.1109/CCPR.2010.5659140
