Benchmark Database and Objective Metric for Wide-Field-of-View Light Field Image Quality Measurement

Light Field Images (LFIs) capture both angular and spatial information, providing a richer representation of scenes. However, due to the limitations of commercial light field camera hardware, the captured field of view is relatively narrow. To address this, image stitching techniques are commonly employed to extend the field of view, resulting in Wide field of view Light Field Images (WLFIs). Given the massive data volume of WLFIs, various visual compression distortions are easily introduced during compression, posing new challenges for measuring the quality of distorted WLFIs. To tackle this issue, this paper first conducts a subjective experiment to measure the quality of distorted WLFIs and constructs a new benchmark quality measurement database, which includes corresponding subjective quality scores. Specifically, each WLFI in the database is generated by stitching 13 LFIs captured by a light field camera, with the focus on distortions introduced by mainstream light field image compression techniques. Furthermore, a novel objective blind metric based on epipolar plane pseudo video sequence representation is proposed to measure the visual quality degradation of WLFIs. Finally, comprehensive experiments are conducted on the proposed database, comparing the proposed metric with several state-of-the-art light field image quality assessment metrics. Experimental results demonstrate that the proposed metric outperforms most existing metrics in various aspects. The proposed database and metric will be publicly released at https://github.com/oldblackfish/NBU-WLFI.