No-reference video quality assessment from artifacts and content characteristics: a neuro-fuzzy framework for video quality evaluation

Reliable assessment of video quality is important for video service providers as it directly impacts a viewer's quality of experience (QoE). Proliferation of video-based applications necessitates accurate and realistic video quality assessment. Impact on a user's QoE can be estimated by evaluating the perceptual quality of a video. In this direction, two no-reference video quality assessment models are proposed in this paper. The first one is feature extraction based predicted video quality measure and and the second one is dimension adaptation based predicted video quality measure. These two predict the quality scores of a distorted video respectively before and after adaptation of its encoding dimensions. The models compute the initial frame level quality scores using a set of spatio-temporal distortions and content characteristics extracted from the video. Afterwards, a multistage temporal pooling mechanism transforms the frame level quality scores of a video into video level quality scores. Finally, a trained neuro-fuzzy model predicts a video quality score for the video. The performance of each of the models is evaluated while predicting quality of videos in publicly available video quality databases. It is observed that the predicted quality scores have good correlation with the corresponding subjective quality scores.