A Proposed Algorithm for Shot Boundary Detection Using Various Features

Abstract

The Shot is often used as a basic meaningful unit for analysing, summarization and indexing video. In this paper, an effective algorithm is proposed to detect abrupt and gradual shot boundary automatically. The proposed algorithm consists of three main phases: fade in/out detector, abrupt detector, and gradual detector. Fade in/out detector is handled in the first step of the proposed algorithm using Maximally Stable Extremal Regions (MSER) descriptors to avoid misdiagnosis caused by black frame. Second, abrupt transitions are detected by analysis of gist features similarity and adaptive local threshold. Lastly, gradual transitions are found out between neighbouring abrupt using gist features as introductory detection of the gradual shot while the color histogram features are employed for finishing detection to reduce the false detection. Experimental results indicate that the proposed algorithm provides 0.981 F-measure, 0.977 precision and 0.986 recall rates for selected testing videos. The performance of the proposed algorithm is satisfactory for detecting different models of transitions under motion effects and camera operations.