Abstract:
The human visual system is attracted to unusual or unique objects in a visual scene. In case of computer vision , Visual Saliency is defined as the property that makes certain objects or regions to stand out in an image or scene . Rapid advancement in technology involves video consumption in various domains. Online streaming, virtual reality and video conferencing are some of the glimpses into era of videos. Thus Video Quality Assessment is crucial in these domains. Visual Saliency plays a vital role in video quality assessment by targeting visually important video regions. This helps in the development of more robust methods for evaluating video quality, streaming protocols and video compression algorithms.
Moreover, it significantly reduces the complexity of evaluation , time and resources. This research work utilizes Global contrast-based visual saliency for video quality assessment . Four methodologies are proposed: Feature Extraction, Feature Selection, Feature Extraction combined with Feature Selection, and Visual Saliency combined with Feature Extraction and Feature Selection. LIVE Video Quality Challenge Database is utilized to evaluate the proposed methodologies. The results of the proposed methodologies indicate remarkable performance when compared with state-of-the-art methods. This indicates that visual saliency and feature selection play a significant role in improving the video quality assessment, while efficiently decreasing the computational time.
