Subjective and Objective Quality of Experience of Free Viewpoint Videos – Paper Overview

IEEE Transactions on Image Processing (IEEE TIP) is an A‑class journal recommended by the China Computer Federation (CCF) for the computer graphics and multimedia field. It focuses on image processing, imaging systems, image scanning, display, and related signal‑processing topics.

The paper titled Subjective and Objective Quality of Experience of Free Viewpoint Videos investigates the quality assessment of Free Viewpoint Video (FVV), a novel immersive viewing mode that allows users to freely adjust the viewing angle in a 360° environment. The authors’ lab at Youku (Alibaba Group) has developed a self‑built FVV interactive video technology that reconstructs and renders dynamic 3D scenes, enabling users to drag and change viewpoints directly in the Youku app.

To quantitatively evaluate the interactive experience, the authors conducted a large‑scale subjective‑objective study in real Youku application scenarios. They proposed a data‑collection strategy guided by intrinsic influencing factors and limited scenes, and introduced a two‑stage coarse‑to‑fine subjective labeling workflow that reduces labeling cost by about 17%.

For the objective side, a benchmark model based on sparse‑sampling of video frames is presented. Experiments show that an appropriately sized sparse frame sequence can still capture the quality variations of FVV effectively.

The collected dataset and the source code have been publicly released at https://github.com/QTJiebin/FVV_QoE .

Authors: Yan Jiebin (Alibaba Group, Jiangxi University of Finance and Economics), Li Jing (Alibaba Group), Fang Yuming (Jiangxi University of Finance and Economics), Che Zhaohui (Alibaba Group), Xue Xia (Jiangxi University of Finance and Economics), Yang Liu (SANY Heavy Industry Co. Ltd). The work mainly originates from Alibaba Entertainment Moku Lab and Professor Fang Yuming’s team at Jiangxi University of Finance and Economics.

Alibaba Entertainment Moku Lab is the core AI technology department of Alibaba Entertainment, focusing on computer vision, graphics, NLP, and other AI areas for the entertainment industry. Its achievements include video quality assessment, intelligent video creation, multimodal tagging, content‑ID systems, interactive effects, and free viewpoint video technologies.