基于自动确定特征点数的镜头分割方法

镜头分割对实现基于图像内容的视频检索具有重要作用。为了提高视频镜头分割的效果，采用Harris 算子进行像素点的能量评估，并抑制伪特征点。通过图像帧分块和边缘检测，获取最小化特征点集，并使特征点分布相对均匀合理，然后根据特征点计算图像帧统计量，进一步形成图像帧的特征向量。在此基础上，结合图像帧渐变检测和闪光检测实现镜头分割方法，取得很好镜头分割效果。

Approach to Shot Segmentation Based on Automatic Decision ofFeature-Point Number

Shot segmentation is very important for video retrieval based on image content. Inorder to improve the performance of shot segmentation, Harris operator is adopted to evaluate theenergy of image pix, and exclude pseudo feature points. By means of partition of frame in thevideo into blocks and edge detection of the blocks, the minimal feature-point set is obtained andthe feature points are made relatively well-distributed. And then according to the feature points,statistical quantity of frame in video is computed to construct feature vector to express the frame.Based on above, along with gradual transition detection and flash detection of frame, a method ofshot segmentation is presented. Experiments of shot segmentation are efficient.