Action based Video Summarization

Abstract

The enormous growth of digital multimedia has resulted in increased complexity of time and data management. Many researchers have developed techniques for managing this tremendous amount of data or longer videos effectively. Video summarization is one of the most important research areas for effective comprehension of video content by selecting the informative frames of the video as representatives. This paper presents two-step action based video summarization approach. First, video frames are extracted and content based key frames are selected. Second, video summary is generated by removing redundant frames by employing feature extraction and classification technique. Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) methods are used for feature extraction. Convolution Neural Network (CNN) is used for classification of frames and video summary is generated using $K$ -Means clustering and Bayesian model. Experiments are conducted on SumMe [1] video dataset by varying feature extraction, classification and summarization methods. Summary of the video is compared with summarized video time and number of reduced frames. Summary is generated by retaining key features of the video.