Object tracking using level set and MPEG 7 color features

This paper investigates the use of a binary level set-based approach for tracking vehicles in video sequences captured by camera systems. A combination of color and segmentation features including background subtraction and connected component labeling will be incorporated into the level set-based framework to achieve the tracking. Next, the retrieval task with respect to a given user's request image is achieved using a set of similarity indices constructed through MPEG-7 color features; namely, color histogram, dominant color descriptor and color layout descriptor. The performances of the system are evaluated in terms of precision and recall evaluations.