Abstract: The authors presented a neural networks based approach of shot boundary detection by using multi-video features. Two approaches, based on feature differences between two adjacent frames and shifting window respectively, were employed to detect abrupt transition, and motion information was used to reduce the influence of strong movement of objects. The fusion and voting techniques were exploited in the final decision stage. In gradual change detection, three patterns of the variance curve of intensity during the dissolving period were distinguished using three neural networks respectively. Then the interference was eliminated according to the characteristics of linear increasing or decreasing of the mean value of intensity during dissolving interval. Experimental results on TRECVID database indicate that the proposed approach works well in detecting shot boundary measured by both recall and precision, and it is also robust to motion and flash light.

Key words: Shot Boundary Detection (SBD)

摘要： 基于神经网络的机器学习思想，提出一种利用多种视频特征的镜头边界检测算法。突变检测中，在特征矢量形成上，分别采用了相邻两帧差值法和滑动窗口法，并加入运动信息以排除强运动对突变检测的影响；在神经网络的构架上，则分别采用了融合法和选举法。在渐变检测中，先通过三个神经网络将溶解过程中方差曲线的三种模式分别识别出来，再根据溶解过程中亮度均值呈线性递增或递减的特性将干扰排除。对大量TRECVID视频进行实验的结果表明，该算法对视频突变和渐变都具有良好的检测性能，并对运动以及闪光灯的干扰具有较好的鲁棒性。

关键词: 镜头边界检测, 视频分割, BP神经网络, video segmentation, Back Propagation (BP) neural network