Mutual reference frame-quality assessment for first-person videos
Chen Bai, Amy R. Reibman · 2017
First-person videos (FPVs) captured by wearable cameras are explored for applications of sharing experiences, recording daily lives, measuring social interactions and behaviors. These applications can be improved by an accurate quality assessment. To maximally use the information present in a FPV, we introduce a new strategy for image quality assessment, called mutual reference (MR). MR does not fit into the previous categorization of full-reference, reduced-reference and no-reference. It uses the overlapping content between images to provide effective information for quality estimation. We propose a framework of mutual reference frame-quality assessment for FPVs (MRFQAFPV) to implement the MR strategy based on a MR quality estimator (QE), LVI. The effectiveness of MRFQAFPV is demonstrated in a subjective test by comparing with 3 no-reference QEs and frame-to-frame motion.