Motion Trajectory Based Spatial-Temporal Degradation Measurement for Video Quality Assessment

Jinjian Wu,Yongxu Liu,Guangming Shi
DOI: https://doi.org/10.1109/vcip.2018.8698612
2018-01-01
Abstract:With the rapid growth of digital video through the Internet, a reliable video quality assessment (VQA) technology is greatly demanded for video management. Motion information plays a dominant role for video perception, however it is too difficult to be accurately analyzed for VQA. The human visual system (HVS) is highly adaptive to track moving objects with pursuit eye movement. Inspired by the motion process in the HVS, we suggest to analyze the degradation along attended motion trajectories for VQA. As a convenient representation of motion, optical flow is calculated for motion trajectory searching. Next, the degradation on the motion velocity along each trajectory is analyzed with the optical flow for temporal quality measurement. Meanwhile, visual information along each trajectory is extracted for joint spatial-temporal quality measurement. Finally, considering the spatial quality degradation from each frame, a novel VQA model is introduced. Experimental results on the public available VQA databases demonstrate that the proposed VQA model performs highly consistency with the subjective perception.
What problem does this paper attempt to address?