Tracking Moving Objects Using Bandpass Filter Enhanced Localisation and Automated Initialisation of Active Contour Snakes

This paper describes a robust and accurate method of Active Contours, also known as Snakes. Active Contours (AC) are a completely different approach to feature extraction. The feature being extracted is the contour of an object in an image or a video frame. Active Contours can also be used to allow tracking of any object in a video sequence using the knowledge of the contour of the object being tracked. In this paper a new method of minimization of boundary co-ordinates of the object is used for the Snake initialization. The boundary co-ordinates of the object under study are extracted using threshold, filtering and morphological operations. A fine tuned Difference of Gaussian (DoG) filter is used to extract the edge image of the object in a scene. The DoG filtered image is also used to add a new energy functional term in the AC Snake energy equation. A new method of adding and removing snake points based on the distance between the points is introduced. Thus, in this paper, a modified energy functional for energy minimization and a new method of AC Snake initialization is proposed. The new snake model designed is dynamic and can be applied both on static and dynamic video frames. In order to show the performance of the proposed snake model several experimental results are presented for frame data from video sequences.