A Video Transcoding Scheme for E-Learning Multimedia Applications

In this paper, we propose a segmentation based transcoding scheme for adapting MPEG-2 e-learning visual contents to heterogeneous environments. This is achieved by converting MPEG-2 video into MPEG-4 video objects with arbitrary shape and different semantic value in e-learning context. The transcoding scheme is based on a hybrid segmentation method, which employs both compressed and pixel domain techniques, for extraction of two video objects from MPEG-2 streams. The objective is two-fold: i) to enable individual object coding and manipulation; ii) to increase the scene coding efficiency. The results show that our hybrid segmentation method is capable of identifying the video objects of interest with good accuracy. Moreover, the transcoding efficiency of the proposed scheme is better than straightforward conversion from MPEG-2 to MPEG-4.