A parallel context model for level information in CABAC

Context-adaptive binary arithmetic coding (CABAC) is one of the most time-consuming modules in H.264/AVC decoder. A potential way to accelerate CABAC is by parallelization. However, the context modeling process for level information in CABAC is highly serial in nature and can not be parallelized in the coefficient level. In order to improve the throughput of CABAC, in this paper we present a parallel context model for level information. The key feature of the model is to use the total number of the significant coefficients and the scanned position of the current significant coefficient in the quantized transform coefficient block as the context information. Since the context information is independent of the previously decoded significant coefficients, parallel decoding in coefficient level is achieved. In experiments, the proposed context model achieves the similar compression efficiency as the CABAC.