Research on Parallel Rate Control Based on BP Neural Network

As an important part in the video coding technology, the rate control aims to achieve a smooth output of the code stream and adjust the quality of the video coding under a limited bandwidth. During the parallel coding framework, the existing rate control algorithm has its limitation, that is, for the I frame, the bit prediction error is large. This drawback will lead to the result that the quality of the video coding cannot be guaranteed. Since back propagation(BP) neural network shows better performance on approximating complex nonlinear relationship and possessing the advantages of high fault tolerance, strong self-learning ability and controllable prediction error, we propose a new rate control algorithm based on BP neural network for parallel coding framework in this paper. Experimental results demonstrate that the frame-level bit prediction error based on BP neural network is smaller and more stable than the traditional method, which verifies the accuracy and feasibility of this method.