Genetic algorithm for inter-frame region object temporal correlation in binary partition tree

Video contains rich information in the spatial and temporal domains. A single video shoot consists of a number of frames which are generally composed of similar objects which may have changed in position and size. The ability to recognize regions in the current frame is an important task in video analysis, but their subsequent recognition in the next frame receives little attention. This research utilizes a region based binary partition tree (BPT) as the content representation; object searching is conducted inside the binary tree and not in the original pixel domain. This work was conceived from the postulate: If an object exists inside a binary tree partition for a given video frame, can a corresponding branch be found in the BPT of the next frame? This is a difficult problem with a one to many and a many to one mapping and requires an innovative solution in the form of a Genetic algorithms (GA). GA's are ideally suited to this problem as the start and end conditions are known. This research proposes and achieves temporal correlation in region based BPTs using a genetic algorithm.