Modelling brightness perception for high dynamic range television

High dynamic range (HDR) television offers greater contrast and more immersive images than conventional television, and as such it is an important part of the overall ultra-high definition television package. Standardisation is almost complete, and the industry is already taking the first steps in HDR programme production. The extended dynamic range and brighter screens associated with HDR make sudden jumps in brightness possible. To ensure consistency between programmes and to avoid uncomfortable brightness shifts at programme junctions, some production guidelines are needed for HDR brightness, just as guidelines have been necessary for audio loudness. In order to develop production guidelines for brightness, brightness perception must be understood. In this paper we report the results of subjective tests that measured the overall perceived brightness of a set of HDR images. We then propose ten classes of potential objective metric that relate the displayed pixel luminance levels to the overall subjective brightness level, and evaluate them using our test results as ground truth. The most effective metrics tested are the mean of the pixel luminances, the mean of the pixel luminances raised to the power of 0.82, and the 96th percentile of pixel luminances, all of which performed similarly well. The mean displayed pixel luminance is preferred, since it is the simplest to implement. The effectiveness of these simple objective metrics suggests that real-time brightness monitoring in production is a realistic goal.

I. Introduction

High Dynamic Range (HDR) television [1] increases the constrast range of television images, and so can exploit the capabilities of ever-brighter consumer displays to create a more natural and immersive viewing experience in the home. Standards for end-to-end television production and distribution are nearing completion, HDR televisions are available in the shops, and the first HDR programmes are being produced.