2D Image Rendering for 3D Holoscopic Content using Disparity- Assisted Patch Blending

Holoscopy is a 3D technology that targets solving some of the limitations of current 3D technology, such as the need to wear special glasses to get the depth perception and the visual discomfort caused inherent accommodation issues. The display technologies available today at home are not yet compatible with 3D holoscopic image formats, for this reason methods that convert 3D holoscopic formats into 2D and current generation 3D image formats are required. There is, however, no fully automated, all-in-Focus and high performance 2D extraction method available in the literature. Moreover, the development of this type of conversion technology requires appropriate testing and assessment. Besides subjective testing with real people, there is nowadays no other reliable way to assess the performance of this type of conversion technology. In this context, and after making an extensive review of 3D holoscopic capture and display technologies, this paper: a) proposes a novel, fully automated, 2D image extraction method for 3D holoscopic images and assesses its performance; and b) identifies potential No-Reference Image Quality Assessment metrics able to rate this type of 2D extractions and assesses their correlation with the human perception of quality. The proposed 2D extraction method – Disparity-Assisted Patch Blending – outperforms, in “normal” conditions, all the available alternative methods. The NIQE metric seems promising as a potential candidate to reliably rate 2D extractions, although more research work should be done on BRISQUE as well to make it even more reliable.