HDR schemes and the issues governing quality color conversions (Part 2)
Anyone attending IBC this year will have been made aware that the addition of High Dynamic Range (HDR) is now regarded as the key parameter to bolster the chances for a UHD channel introduction to complement existing HD services (see part 1 All aboard the HDR Tram). The principle reason for the introduction of HDR is to give the format a clear, definable picture quality advantage over HD. In no way are broadcasters yet contemplating a like for like swap out from HD to UHD, far from it.
A key concern is whether UHD acquisition can eventually be used as the source for crucial revenue generating HD services. The issue at hand is the quality of the UHD HDR to HD SDR conversion, which is effectively three conversions in one. Firstly HDR to SDR, as well as UHD to HD, and then the implied conversion from the 2020 color plane to 709, since all of the standardization effort for UHD is behind the expanded color gamut offered by 2020. While the inclusion of HDR in the mix does complicate the necessary conversion, making it a true color space conversion rather than just a color plane, it was clearly the move from 2020 to 709 and the potential for hue shifts and color flips that concerned visitors to IBC this year.
The most severe effect from this conversion was poor color portrayal in the resulting HD pictures when compared to the UHD HDR source. This aspect was a key reason why the end-to-end workflow, shown on the Harmonic booth, demonstrated that it was possible to produce an HD SDR feed suitable for streaming, but not a pristine HD broadcast stream from a UHD HDR source. There is a long way to go before ruling out the prospect of a quality conversion, suffice to say the various converters demonstrated at IBC demonstrated problems deriving an SDR HD feed from a UHD HDR source. Cracking this issue is fundamental to the economics associated with UHD channel launches, particularly for live event coverage.
Some show attendees concluded that moving from 2020 to 709 introduces unacceptable color shifts, particularly at the gamut edges. The only way forward for those in this camp is to consider the more radical option of back-to-basics reallocation of bits. Inevitably, fundamental problems with artifacts introduced due to non-linear function are under scrutiny here. Basically the end-to-end chain is a compromise when it comes to chrominance down-conversion. This inevitably turns attention to the suitability of 2020 as a basis for future UHD, especially when compression for channel distribution is considered.
To most, the prospect of fundamental re-evaluation of color science behind UHD delivery is not up for discussion, given that screens are already reaching expectant viewers in massive numbers based on 2020 color. For these realists, discussion at IBC was limited to bit depth, content bandwidth needs and the practicality of the metadata schemes associated with the numerous HDR schemes. Even the basis of metadata was debated. Many attendees were keen to examine the downstream signaling describing the scene and capture environments that ultimately populate the crucial metadata fields and how critical the timing is in the case of schemes requiring dynamic metadata.
Harmonic is analogous to Switzerland when it comes HDR, ultimately we do what our customers ask for. This makes it sound like we neutrally sit on the fence, far from it. Those deploying UHD channels are genuinely in a quandary concerning deployment, particularly from a workflow perspective. In showing a complete live end-to-end workflow at IBC, Harmonic attracted many who wanted to see HDR in context, i.e. native UHD throughout, not 1080P, and on representative screens that are consumer rather than professional grading quality. Sure, a 4000 nit grading monitor in a darkened suite provides ample scope for highlighting chrominance down conversion problems for those fundamental to the various HDR schemes, but for the vast majority of attendees getting a handle of the issues from customer perspective was critical. To that end, using 65” LG OLEDs under harsh lighting, showing the complete contribution, play-out and distribution workflow for a static HDR metadata scenario caught the imagination of those attending the booth.
There is still scope for improvement with this workflow scenario, as many attending the show wanted to see a mixture of HDR schemes handled at ingest by a system able to handle and interpret the various signaling/metadata approaches. Diversity in terms of screen types showing HDR content is also another key worry, since most of the graded content was effectively being demonstrated as canned footage graded on specific screens and not material graded that can be successfully mapped across all screens. These two areas will no doubt keep me occupied during the down time between IBC and NAB. In part 3 of my post-IBC blog series, I’ll move away from UHD and tackle the key issue behind today’s infrastructure, namely virtualization, COTS and SaaS, i.e. where is video processing heading?
– Ian Trow, Senior Director of Emerging Technology & Strategy at Harmonic