Video Encoder Implementation

Abstract—In this paper we extend a previous encoding latency analysis framework for arbitrary multiview video coding pre-diction structures with a new multi-processor multiview encoder model. As opposed to the fixed scheme where each processor is in charge of encoding a single view, this new model focuses on a flexible assignment of the encoding of frames among the pool of available processors. This new model is more suitable for realistic scenarios in which the number of views and processors is different. Moreover, we show that for this new encoder model, a more efficient use of the processors can be achieved by means of appropriate frame coding ordering algorithms. This is especially interesting in real-time encoding applications in which frame processing time constraints are imposed in order to maintain low encoding latency. Finally, we evaluate the latency performance of this new encoder model and show its advantages over the fixed encoder scheme. I