Ribera

Abstract

The wide field of view in head-mounted displays implies that most of the content shown by the display appears at high eccentricities. However, most image quality metrics assume that the image is viewed by the fovea, where the human visual system presents its highest acuity. In this project, we extend an image quality metric based on ISETBIO to wide field of views. We apply publicly available measurements of the Optical Transfer Function of the human cornea and lens at different eccentricities. In this manner, our Delta E error map becomes eccentricity-dependent. After incorporating more accurate OTF models to our method, we expect to use this image quality metric as a tool to design newer displays technologies that take advantage of the loss of acuity in higher eccentricities.

Introduction

Head-mounted displays are pushing the display specifications (e.g, resolution or refresh rate) much beyond those of traditional displays. One of the reasons is the need to cover a huge field of view with a higher pixel density, since the resolvable pixel size at such close distances is much smaller. This unavoidably leads to a tremendously higher need for bandwidth in the transmission lines that drive the display. Unfortunately the demand outpaces the speed at which newer transmission lines increase their available bandwidth. In this context, image and video compression plays an important role by decreasing the required bandwidth [1]. In order to ensure that the compression algorithms do not degrade the visual quality of the displayed image, compression engineers use image quality metrics.

However, most image quality metrics assume that the image is viewed by the fovea, where the Human Visual System (HVS) presents its highest acuity. This motivates the need for image quality metrics that take into account a high field of view. Such image quality metrics can become a valuable tool to design display technologies for future Virtual Reality headsets.

Background

This project is based on an image quality metric called ISETBIOLAB proposed in [2], which in turns follows the ISETBio approach of modelling every step in the pipeline of the HVS. An schematic of ISETBIOLAB is shown in Figure 1. First, it computes the radiance of the scene emitted from the display by considering the subpixel arrangement and spectral power distribution of the display primaries. Then the modulation transfer function of the human optics (the optical transfer function) is applied to each wavelength of the scene to obtain the irradiance on the retina, i.e, the optical image (also called retinal image). This process is repeated for the second image (the image being compared with). Finally, both retinal images are converted to XYZ and a color difference metric (such as CIEDE1976) is applied pixelwise. This creates a Delta E error map between the original and reconstructed image. Figure 2 shows an example of an image going through ISETBIOLAB.

Figure 1. Schematic of ISETBIOLAB

Figure 2. Example of an image going through ISETBIOLAB

On the other hand, the Optical Transfer Function (OTF) of the HVS has been measured in a variety of ways, including double-pass interferomic measurements [4, 5] and wavefront sensors [6]. These OTFs depend on the eccentricity of the incoming rays crossing the human pupil, thus

References

[1] F. G. Walls and A. S. MacInnis, "VESA Display Stream Compression for Television and Cinema Applications," IEEE Journal on Emerging and Selected Topics in Circuits and Systems, vol. 6, no. 4, pp. 460-470, December 2016. https://doi.org/10.1109/JETCAS.2016.2602009 [2] G. W. Cook, J. Ribera, D. Stolitzka and W. Xiong, "A Subpixel‐based Objective Image Quality Metric with Application to Visually Lossless Image Compression Evaluation," Society for Information Display Digest of Technical Papers, vol. 49, no. 1, May 2018. https://doi.org/10.1002/sdtp.12103 [3] https://github.com/isetbio/isetbio [4] R. Navarro, P. Artal, and D. R. Williams, "Modulation transfer of the human eye as a function of retinal eccentricity," vol. 10, no. 2, pp. 201-212, 1993. https://doi.org/10.1364/JOSAA.10.000201 [5] D. R. Williams, P. Artal, R. Navarro, M. J. McMahon and D. H. Brainard, "Off-axis optical quality and retinal sampling in the human eye", vol. 36, no. 8, April 1996. https://doi.org/10.1016/0042-6989(95)00182-4 [6] B. Jaeken and P. Artal, "Optical Quality of Emmetropic and Myopic Eyes in the Periphery Measured with High-Angular Resolution," Journal of Visual Psychophysics and Physiological Optics, vol. 53, no. 7, June 2012. https://doi.org/10.1167/iovs.11-8993