Detecting Important Video Shots Using Video Summarization and Feature Extraction Techniques

Video Shot Detection is one of the techniques used in detecting changes in video scenes. Shots can be defined as the fundamental units of analyzing video files in a variety of applications, including online advertisements and e-commerce. The important shot determines how much a shot influences the video analysis process. Because of their low substance, some video shots should be discarded in order to reduce video raw data, using the summarization technique. Video summarization provides frames that are of use in creating conceptual information about a certain video. The present work proposes the coordination of both concepts: shot detecting and video summarizing for obtaining better outcomes. Feature extraction is performed via the frames so as to draw a comparison with consecutive frames used in video shot detection. This process involves the use of histogram feature algorithm, Histogram of Oriented Gradients (HOG) and Expectation Maximization (EM), after which video summarization is performed to find the best frame shot using Euclidian distance between two neighboring frames. The shot frames obtained during video shot detection are used in the video summarizing process in order to find or detect the important shots in a video. The obtained results indicate that the method introduced in this work is robust and sufficient.