Assessing the quality of experience in viewing rendered decompressed light fields

Abstract

A light field image is a sampling of the intensity and direction information of the light rays crossing the main lens of a digital light field camera. The light field information captured by a camera must be processed in order to render images of the scene to the final user at a given focal plane or viewpoint. In the area of light field imaging, efficient representation, processing, compression and quality evaluation techniques and methodologies are currently under research in order to foster the development of novel industrial, entertainment or scientific applications. This paper focuses on the problem of evaluating the quality of experience when viewing rendered decompressed light field images. A processing chain for coding and decoding a light field image is first defined as reference model. Then, a novel metric for quality evaluation of the rendered views is proposed. This metric measures the variation of structural similarity on a set of viewpoints extracted from the light field. Subjective evaluation is performed and the correlation between objective metrics and the subjective results is reported and discussed. The proposed objective quality metric resulted in a nearly strong correlation with the subjective assessment results.