A video database for intelligent video authentication

Abstract

In this paper we describe a unique video database which consists of the real life moments of people and objects, captured under various illumination conditions and camera positions. We have classified all the videos of our database into six categories, out of which four categories are based on the movements of camera and objects (captured by the camera). The remaining categories of the database are daylight videos and night vision videos. The videos captured under the natural light source (such as sunlight) are covered in daylight videos category. The night vision videos category has the same setup and environment as in the daylight videos category but the videos are captured in low light condition and the camera is recording in night vision mode. Each category of this video database offers a good situation for the challenge of video authentication and to fathom the credibility of video authentication algorithms too. We have applied our own intelligent video authentication algorithm on each category of the video database and obtain the results with the overall accuracy of 94.85%, subjected to various tampering attacks.