By David Touze, Technicolor; Leon van de Kerkhof, Philips
The arrival of the High Efficiency Video Coding (HEVC) standard enables the deployment of new video services with enhanced viewing experience, such as Ultra HD broadcast services. In addition to an increased spatial resolution, Ultra HD will bring a wider color gamut (WCG) and a higher dynamic range (HDR) than the standard dynamic range (SDR) HD-TV currently deployed. Increasing of dynamic range, i.e. the luminance ratio of bright over dark pixels, has been shown to dramatically improve the user experience. Increasing gamut and dynamic range are two faces of the same coin as they basically augment the color volume to which pixels belong. Furthermore, luminance and colors are intrinsically linked in legacy workflows that are non-constant luminance: the signal non-linearity is not applied directly to the luminance, but instead the non-linear luminance is a combination of non-linear quantities (typically RGB).
Different solutions for representing and coding HDR/WCG video have been proposed [1][2] [3][4]. As stated in [5][6][7][8]. SDR backward compatibility with decoding and rendering devices is an important feature in video distribution systems, such as broadcasting or multicasting systems. The coming American broadcast standard ATSC 3.0 is expected to emit both SDR BT.709/2020 and HDR BT.2020 streams. The European DVB standard has already introduced SDR UHDTV in the BT.2020 color space and will extend it to HDR BT.2020 soon. Peak brightness is expected to migrate from legacy 100 nits to about 1000 nits, but compression solutions should be flexible enough to handle future higher brightness as well as non-broadcast applications that may take advantage of more nits.
Dual-layer coding, for instance using the scalable extension of HEVC (a.k.a. SHVC) is one solution to support SDR backward compatibility. However, due to its multi-layer design, this solution is not adapted to all distribution workflows. An alternative is to transmit HDR content and to apply at the receiving device an HDR-to-SDR adaptation process (tone mapping). One issue in this scenario is that the tone mapped content may be out of control of the content provider or creator. Another issue is that a new HDR-capable receiving device is needed to apply this tone mapping for existing SDR displays.
Alternatively, the Hybrid Log Gamma (HLG) transfer function [2] has been designed as a straightforward solution to address the SDR backward compatibility, that is, an HDR video graded on a display using the HLG transfer function can be in principle directly displayed on an SDR display (using the BT.1886 transfer function [2]) without any adaptation. However, this solution may result in color shifting when the HLG-graded video is displayed on an SDR rendering device, especially when dealing with content with high dynamic range and peak luminance [10][11][12]. Also, there is no way to optimize the brightness and contrast of the SDR image.
The proposed Single Layer SDR backward compatible HDR video distribution solution detailed in this paper, named SL-HDR1, and standardized in ETSI TS 103 433 specification [13], aims at addressing these issues. SL-HDR1 leverages SDR distribution networks and services already in place. It enables both high quality HDR rendering on HDR-enabled CE devices, while also offering high quality SDR rendering on SDR CE devices.
The main features of the HDR distribution system are as follows:
- Single layer with metadata: SL-HDR1 is based on a single layer coding process, with side metadata that can be used at post-processing stage. The metadata payload corresponds to a few bytes per picture, GOP or scene.
-
Codec agnostic: SL-HDR1 does not impact the core codec technology and is codec independent. SL-HDR1 is based on an encoding pre-processing applied to the HDR input, and on a corresponding decoding post-processing (functional inverse of the pre-processing) applied to there constructed video from decoding. Use of a 10-bit codec is recommended, since an 8-bit codec could introduce artefacts such as banding effects, due to having too few code words available for the precision required for HDR content.
-
Enable SDR backward compatibility: a decoded bitstream can be displayed as is on an SDR display. The color fidelity is preserved compared to the HDR version. An additional post processingis applied to convert the decoded SDR version to HDR, thanks to the metadata, with preservation of the HDR artistic intent.
-
Enable preserved quality of HDR content: there is no penalty due to the SDR backward compatibility feature; coding performance compared to HLG are improved, in particular in terms of color impairments.
-
Enable adaptation of the HDR content to the HDR display capabilities: if the HDR content peak brightness is higher than the HDR display peak brightness, the post-processing adapts the HDR content to display peak brightness, preserving all details and HDR artistic intent.
-
Limited additional complexity: the pre- and post-processing steps are of limited added complexity; in particular the involved operations are only sample-based, without inter-sample dependency.
-
Independent from the input OETF: the pre- and post-processing operate in linear-light domain, and are therefore independent from the input OETF.
The document is organized as follows. The solution overview is presented in section 1. Section 2 describes the HDR-to-SDR decomposition and section 3 the HDR reconstruction process. Section 4 relates to the metadata signaling. Section 5 details the display adaptation feature. Section 6 presents tests results, assessing the SDR quality and the HDR compression performance of SL-HDR1 comparatively to distribution solutions based on PQ and HLG transfer functions. Conclusion section provides closing remarks.