June 16, 2015, SMPTE Entertainment Technology in the Internet Age, Stanford, CA—Lars Borg from Adobe describes some of the differences between standard dynamic range and high dynamic range workflows. These new specifications calling for wider color gamut and higher dynamic range will impact content workflows.
Current HD TV is based on BT 709 to provide 1920 x 1080 images with a slim color gamut. The first generation UHD or 4K sets increase resolution 2840 x 2160 but retain the BT 709 color gamut. The next generation of 4K sets and the 8K will use a much wider Rec. 2020 color gamut. At the same time, standards bodies are considering higher dynamic range for these TV sets.
The first difference between generation one generation to systems is that the acquisition phase has no HDR perceptual quality (PQ) capabilities built in. The data streams will use the same SDI packaging as 4K, 12 bit but there is no current capability to include PQ flags. The raw streams of compressed through H.264 or H.265 for transfer to other processing functions. H.265 does include a flag for PQ.
For in-camera color grading, HDR has a much larger grading latitude and is therefore not the same as SDR system gamma. The on-set 3-D out could be HDR PQ and final grading processed in post. This processing is currently available within the existing law encoding to get 12 bits from the 16 in the cameras. The computer generated images and digital visual effects don't need to consider these conversions since their outputs can be determined and set at render.
The biggest challenges are that some TV sets are still operating in ST including standard dynamic range. The conversion between standard dynamic range of high dynamic range video must consider the significant differences in color volume as well as the greatly increased dynamic range. Converting SDR to HDR using Rec.709 leads to lost out colors and lost coding range. The alternative is to use BT 1886 for closer approximations and fidelity to the Rec. 2020 standards.
Color grating in SDR depends on the media space and gamma which is a function from 0 to 1. This is not good if the content will be repurposed for HDR since a lot of information is lost in moving to Rec 709. It is better to work in an HDR-native workflow that includes conversion capabilities to SD and Rec 709 as an output, and to other encoding spaces. The new PQ standards and recommendations combine the characteristics of gamma at low luminance levels and a log function at higher levels.
In the cinematic areas, the Academy Color Encoding System (ACES) is HDR ready and includes an intermediate 16-bit floating-point mezzanine representation. The real problem is that the distribution standards for HDR aren't ready yet. A direct conversion to standard dynamic range leads to crushed blacks and color shifts. These characteristics degrade presence details and local contrast. They also create artifacts like halos, unreal appearance, and slow visual analysis.
SMPTE is moving towards a single master and multiple targets for video files. These systems will call for color metadata for color transfers on a per frame basis. The challenges are determining how to embed these data into the video frame information, since the transforms can be put into effect at the display with the metadata.
