Scene transition based adaptive GOP selection for increasing coding efficiency & resiliency

In hybrid video coding, raw video is encoded exploiting statistical, temporal, and psycho-visual redundancy. Due to its predictive nature, error accumulates cumulatively, as exact prediction is not possible. Therefore, after a certain interval, it is required to refresh the prediction model to stop error propagation in subsequent frames. These frames can be grouped to form a Group of Pictures (GOP). In this paper, we propose a novel algorithm to fix Group of Picture (GOP) size adaptively depending on the scene content. The proposed method works on the extent of difference of a frame with respect to the reference frame. The threshold adopted by us remains constant over a range of data set having different motions and adaptive decision is taken online during the encoding process. We also propose a channel coding scheme to make the compressed bit stream error resilient. Thus, our proposed method increased coding efficiency as well as the resiliency measured in terms of PSNR of the reconstructed video over fixed length GOP.