Abstract: H.264/AVC is the newest video coding standard with high compression efficiency.But it will take a long time to code with extremely complex algorithm，it can not achieve real-time application.In this paper，according to the characteristic of intra-frame prediction algorithm，a fast adaptive thresholds algorithm based on MAD is proposed.This algorithm will predict every macro block before coding，then select between intra 4×4（I4） and intra16×16（I16） prediction modes according to the adaptive thresholds.Then，in I4 prediction modes，select the best mode from its nine prediction modes according to the threshold. Therefore，the complexity of algorithm and coding time are reduced dramatically.Experimental results show that the coding time of proposed algorithm is reduced average by 61.3%，with a little bit increase，and the PSNR is invariable.

摘要： 最新的视频压缩标准H.264/AVC具有极高的压缩率，但其算法极其复杂，编码时间较长，无法达到实时应用的要求。针对其帧内预测算法的特点，提出了一种基于MAD的自适应阈值快速帧内预测算法。算法充分利用宏块的MAD（平均绝对误差，Mean Absolute Differences）信息及时空相关性，在进行帧内预测之前先对宏块预判,同时采用自适应阈值的方法在帧内4×4（I4）和帧内16×16（I16）预测模式之间快速进行选择；然后针对I4预测，采用阈值法在9种预测模式间快速选择，从而减少了算法的复杂度，提高了压缩速度。实验结果表明，所提的算法在码率只有少许增加的情况下，编码时间平均减少61.3%，PSNR值基本不变。