February 15, 2019
Sony Pictures chief technology officer Don Eklund presented a look at HDR bias light analysis. Put plainly, the colorist grades the content in a dark room, with the light behind the screen rated at between 5 and 10 nits. But the average viewer watches that same content in a room with windows and lights. “We have a fundamental problem everyone here has experienced, with light leaking in the room through drapes,” said Eklund. “It crushes the blacks and offers colors that were not intended when the images were graded.”
“It’s an easy problem to describe but not to measure or solve after a couple of iterations of testing,” admitted Eklund. He described the work done to attempt to quantify the problem. “The purpose of the experiment was to understand the effects of a bias backlight on perception of an image,” he said, adding that the test was limited to picture contrast and didn’t consider white point (a topic for a future test).
Eklund conducted the test under the auspices of the UHDTV committee of ASC’s Motion Imaging Technology Council (MITC). The test was organized and hosted by Sony Pictures, and the final data was compiled by Eklund, Daniel De La Rosa and Stephen Shapiro.
The test site featured two Sony 4K UHD Z9D TVs positioned at an angle, one set to a backlight of 10 nits and the other to 100 nits. Both were calibrated for color and luminance with a Klein K10A colorimeter. “The first time we did a test, we let the participants adjust lift and gain using a control surface and discovered that there’s a reason why colorists get paid a lot of money,” said Eklund.
The team refined the test, with a colorist creating pre-sets for 15 pre-selected images, based on quality of the image, details in the shadows and blacks and other images representative of a bright daylight scene. Participants could then choose their preferred pre-set. Eklund went through the results of the tests, showing how participants’ choices were not uniform, with some images drawing votes for three different pre-sets.
“It tells you that people aren’t always looking at the same thing and prioritizing the same things in the image or there would have been more consistency,” said Eklund.
“The initial conclusion is that preferences for compensation vary by user,” said Eklund. “In general people prefer to enhance shadow detail in dark scenes. Color saturation is a factor in dark scenes; bias light at 100 nits leads to slightly desaturated appearance in dark scenes. A single compensation curve will not provide desired results for all images.”