The initial justification for a move towards 4K was largely made based on improved resolution. However, frame rate has long been the parameter the experts recommend to give the biggest performance improvement over HD.
A third factor critical to the success of Ultra HD is colorimetry – the science of color perception. Even if the relative importance of these parameters is questionable, what is not up for debate is the fact that colorimetry is sure to be the most complicated parameter to implement being made up of many interrelated issues, if it's adopted for Ultra HD.
As it stands, Rec. 709 (ITU-R specifications for HDTV) color space is being used as the basis for Ultra HD, at least until the industry decides how to handle Rec. 2020 (ITU-R specifications for UHDTV). To the casual observer it may seem strange that Rec.709 is being considered at all for Ultra HD when Rec.2020 has been around since 2012 and is implemented in both the latest version of HDMI (the one we're not supposed to call 2.0) and the latest MPEG compression spec (HEVC or H.265 depending on your perspective).
The lack of adoption is nothing to do with transfer characteristics of Rec. 2020, for 10 bits per sample the non-linear transfer function is identical to that deployed for Rec.709. This largely explains why Rec. 709 can be applied to Ultra HD, so why consider 2020 at all then? To answer this we have to return to the aims of Ultra HD, which is namely to improve on HD and deliver the wow factor associated with color rendition in the cinema. In addition to addressing frame rate and resolution, Ultra HD aims to close the gap between cinema and television by improving the dynamic range of Ultra HD screens to match the capability of cinema.
So how do we do this? One approach is to define a wider color space. Rec. 2020 offers a vast improvement over Rec.709 in this respect. Another possibility would be to extend the number of bits per sample. Rec. 2020 supports both 10 and 12 bits per sample. So let’s take each of these approaches in turn. A wider color space has obvious benefits, especially when Rec. 709 compares poorly with the color space supported by film.
Put simply, Rec. 709 doesn’t have the required scope to support the color space extension hoped for by the industry or indeed possible with the latest generation of Ultra HD screens. Extending the color space does come with problems though. For Rec. 709, a single color space conversion is adopted, for reasons of compatibility with HD screens, but this is far from ideal in a world where Ultra HD screens may be marketed and priced according to dynamic range capability.
For UHD screens to have varying dynamic range capabilities, multiple color space conversions need to be supported according to screen and application type. This greatly increases the complexity and adversely impacts the viability of adoption in media workflows. HDR systems aim to overcome the need for multiple color space conversions by retaining a native color space and only applying tone mapping according to the desired screen. Herein lies the explanation for the difficulty in implementing Rec. 2020, namely what native color space should be adopted and how is the tonal metadata handled?
Lastly, what sample bit depth is really required for UHD? I’d speculate that 10 bit can more than utilize the enhanced color space offered by Rec. 2020 for Ultra HD distribution, provided of course at least 12 bits per sample are utilized further up the production workflow. This is contentious positioning though, so I'll use my next blog to explain my reasoning.
– Ian Trow, Sr. Director, Emerging Technology & Strategy, Harmonic