Abstract: In view of the intra encoding computational complexity of HEVC, a fast intra prediction mode decision algorithm is proposed. Based on statistical data, for prediction units of large size, the algorithm uses threshold method to exclude the intra prediction modes of small possibility. When the size of prediction unit is small, according to the distribution of cost, the algorithm reduces the number of the intra prediction mode directly or uses the method of gradient mode histogram to exclude the intra prediction modes of small possibility, then reduces the number of intra prediction modes for the rate-
distortion cost calculation and cuts down the encoding complexity. The experimental results show that the proposed algorithm can improve the HEVC encoding speed greatly with a negligible image quality degradation and average bit-rate increase, which is in favor of real-time applications.

Key words: intra prediction, statistical results, gradient mode histogram, mode decision

摘要： 针对HEVC帧内预测模式选择过程计算较为复杂的问题，提出了一种帧内预测模式快速选择算法。在统计实验数据的基础上，针对64[×]64、32[×]32、16[×]16较大尺寸的预测单元类型，采用阈值法筛选掉粗选后可能性较小的预测模式，针对8[×]8、4[×]4较小尺寸的预测单元类型，根据代价值的分布情况，直接筛减或者采用模式梯度直方图法筛选掉粗选后可能性较小的预测模式，从而减少最后进行率失真代价精密计算的帧内预测模式数量，降低帧内编码复杂度。实验结果表明，在保证图像质量和编码码率的情况下，该算法能有效提高HEVC的编码速度，有利于实时应用。

关键词: 帧内预测, 统计结果, 梯度模式直方图, 模式选择