Blind video quality assessment based on multilevel video perception

作者：

Highlights：

• This paper proposes a blind VQA model based on multilevel video perceptions (MVP).

• MVP fuses NVS, global motion and motion temporal correlation features.

• Motion compensation filtering enhancement is adopted in MVP.

• MVP is feasible for different types of distortions without knowing distortion types.

• MVP clearly outperforms state-of-art blind VQA metrics.

摘要

•This paper proposes a blind VQA model based on multilevel video perceptions (MVP).•MVP fuses NVS, global motion and motion temporal correlation features.•Motion compensation filtering enhancement is adopted in MVP.•MVP is feasible for different types of distortions without knowing distortion types.•MVP clearly outperforms state-of-art blind VQA metrics.

论文关键词：Blind video quality assessment,Natural video statistics,Motion features,Feature fusion,Video enhancement

论文评审过程：Received 30 June 2020, Revised 4 August 2021, Accepted 6 September 2021, Available online 21 September 2021, Version of Record 25 September 2021.

论文官网地址：https://doi.org/10.1016/j.image.2021.116485