A new video quality metric for compressed video.
Bhat, Abharana Ramdas
MetadataShow full item record
Video compression enables multimedia applications such as mobile video messaging and streaming, video conferencing and more recently online social video interactions to be possible. Since most multimedia applications are meant for the human observer, measuring perceived video quality during the designing and testing of these applications is important. Performance of existing perceptual video quality measurement techniques is limited due to poor correlation with subjective quality and implementation complexity. Therefore, this thesis presents new techniques for measuring perceived quality of compressed multimedia video using computationally simple and efficient algorithms. A new full reference perceptual video quality metric called the MOSp metric for measuring subjective quality of multimedia video sequences compressed using block-based video coding algorithms is developed. The metric predicts subjective quality of compressed video using the mean squared error between original and compressed sequences, and video content. Factors which influence the visibility of compression-induced distortion such as spatial texture masking, temporal masking and cognition, are considered for quantifying video content. The MOSp metric is simple to implement and can be integrated into block-based video coding algorithms for real time quality estimations. Performance results presented for a variety of multimedia content compressed to a large range of bitrates show that the metric has high correlation with subjective quality and performs better than popular video quality metrics. As an application of the MOSp metric to perceptual video coding, a new MOSpbased mode selection algorithm for a H264/AVC video encoder is developed. Results show that, by integrating the MOSp metric into the mode selection process, it is possible to make coding decisions based on estimated visual quality rather than mathematical error measures and to achieve visual quality gain in content that is identified as visually important by the MOSp metric. The novel algorithms developed in this research work are particularly useful for integrating into block based video encoders such as the H264/AVC standard for making real time visual quality estimations and coding decisions based on estimated visual quality rather than the currently used mathematical error measures.