Video Object Extraction Using Model Matching Technique: A Novel Approach

Abstract

Emerging video standards addresses mainly content-based functionalities and interactive multimedia. This requires a prior decomposition of sequences into semantically meaningful physical objects. We formulate this problem as one of separating foreground objects from the background based on motion information. The algorithm begins with the motion estimation accomplished using ARPS (Adaptive Rood Pattern Search) algorithm on any two frames of image sequence. A binary object model representing edge features is derived using Canny edge detector and tracked in subsequent frames using motion vectors generated from ARPS algorithm. The object model is updated for each frame to accommodate for complex motions and shape changes. Motion based region merging is performed finally to merge regions with similar motion to extract moving objects. Experimental results demonstrate the performance of our proposed algorithm.

Video Object Recognition and Modeling by SIFT Matching Optimization

“…Some works use video to detect moving objects by motion. In (Kavitha, 2007), for example, the authors use two consecutive frames to first estimate motion vectors and then they perform edge detection using canny detector. Estimated moving objects are updated with a watershed based transformation and finally merged to prevent over-segmentation.…”