中文摘要：

H.264/AVC QCIF视频场景切换检测是视频应用中一个极具挑战性的研究课题，只利用宏块信息很难得到满意的检测效果，为此提出了一种采用动态阈值和AC相似度的H.264/AVC视频场景切换检测算法，考虑编码预测模式和DCT系数等信息，提出基于编码比特数的动态阈值和基于AC能量的AC图像相似性两个判决准则，据此进行视频场景切换检测。实验结果表明，在高速运动及场景切换频繁的视频上场景切换检测率得到提高，性能也优于现有的算法。

英文摘要：

Scene change detection in H.264/AVC QCIF video sequence is challenging as a difficult task,because the only limited macro block information can not lead to the satisfactory detection accuracy.This paper proposes an effective scene change detection algorithm in H.264/AVC QCIF video sequence.The algorithm is dependent on encoding prediction modes and DCT coefficients which can be extracted from H.264/AVC baseline profile.Two decision criterions,dynamic bit threshold and AC image similarity metric are presented by using the intra- and inter-frame bits and DCT AC energy.Experimental results show that the proposed algorithm can achieve better scene change detection performance for the higher speed motion and the frequently scene change,compared with the related method.