New Indexing method for Content-Based Video Retrieval and Clustering

概要

論文の詳細を見る
An automatic semantic feature extraction of video units is still beyond the current technology of image processing. But, it is possible to extract automatically some low-level features of a video unit. In this paper we introduce a video 'shot' as the basic video unit and propose an idea to automatically extract the color-location features of video shots. The video shot is represented by its color content by means of color histograms. To add the spatial information of color distribution within the frame, video frames are divided into several subregions, that is inspired by the work in [1] which dealt with still images. The average color histograms of all the corresponding subregions in all the frames of a shot are computed, Then, the average histograms' information is used to construct a color-location feature vector which is an abstraction of the representative frame of a shot. In this way we can represent a shot compactly and sufficiently by an abstract frame.
一般社団法人情報処理学会の論文
1997-03-12