Andrey Norkin

HDR video pre-processing

High Dynamic Range (HDR) is considered a major improvement in quality of television pictures. One of the options for streaming HDR video is HDR10 that uses HEVC Main10 encoding, ST.2084 transfer function, BT.2020 color space, Y'CbCr 4:2:0 non-constant luminance color format, and can use some optional supplemental enhancement information (SEI) messages.

A combination of non-constant luminance Y'CbCr 4:2:0 with highly non-linear transfer function can sometimes lead to subjective quality problems. This problem is likely caused by a combination of the steep slope of the opto-electrical transfer function (OETF) in the low-luminance range and the color transform, which make color components with low values have significant impact on Y', Cb, and Cr values. This may cause artifacts in saturated colors at the boundaries of the color gamut.

A fast algorithms that helps to alleviate this problem has been proposed. A closed-form solution optimizing reconstruction in linear R,G,and B space is described here, whereas a solution minimizing a difference in linear luminance value is shown under the following link. The latter solution has been included in an ITU-T report on HDR conversion and coding practices.