Tampered videos-detection and quality assessment

Abstract

In the socio-economic knowledge and scientific development, the images and videos available at various video sharing and social networking websites are playing a significant role and signify their unprecedented role in today’s context, besides the videos captured through CCTV cameras specifically meant for surveillance. On the other hand, advancements in digital video technologies made it easy to manipulate or tamper the videos; hence the authenticity of these videos or videos seen in mass media, such as television, cannot be always taken for granted. In recent years, tampering with video evidences is more prevalent than ever, where facts were misrepresented by the fake or doctored videos. In this thesis we aim to verify the authenticity of a video as well as detection of the location of temporal tampering of frame drop, frame copy, and frame swapping. The proposed schemes neither require any additional bit to embed in videos nor use any coding artifacts. The developed schemes use various thresholds defined with proposed features and achieve good accuracies. Apart from detection of tampering, this thesis also focuses to assess the quality degradation in spatio-temporally and temporally tampered or distorted videos. Based on series of subjective experiments, we have obtained the Mean Opinion Score, MOS and developed a full reference dropped frames quality metric. The assessed quality degradation shows better correlation with the obtained MOS.