In this paper, we we first propose a parallel video coding framework, where three important
factors: parallel strategy, computation complexity and scheduling of tasks are considered. Then
considering the characteristics of medium-grain quality scalability (MGS)-based Scalable Video
Coding (SVC), a new layer-level parallel encoding method is also introduced. Since each encoding task
could have different priority under different parallel strategy, the scheduling of layer-level parallel
encoding tasks is extremely complex. We use Directed Acyclic Graph (DAG) to model the relationship
of encoding tasks and theoretically analyze how to find the optimal solution. Then according to the
DAG model, several heuristic scheduling methods are employed to get a high speed-up ratio.
Experiments show the speedup of our layer-level parallel method was higher than the prior works.
Using the proposed method High Definition (HD) SVC videos can be encoded in real time with SIMD
technology.
