SMPTE: “Trim the UHD codecs”

  •   
  •   
  •   

A paper by a trio of video experts says that video production equipment using different codecs and settings to balance competing requirements, such as trading off between bitrate and quality, and make video production life far too complicated. The introduction of UHD-TV and high dynamic range (HDR) adds to the number of codec implementations, making codec choices harder.

The paper, by Dagmar Driesnack of the IRT, and Simon Thompson and Frans de Jong from the EBU, explains that moving from HD to UHD also means changing several of the key features of mainstream TV systems. First and foremost, the scanning system changes. The traditional interlaced scanning raster allows for a progressive-only system, which can be used at various frame rates up to what is currently seen as high frame rates (100 or 120 Hz including fractional frame rates).

Second, a higher resolution compared to the present HD is also possible. The current focus is on the 2160 lines ×3840 pixels version of UHD, but there is also the 8K resolution (4320×7680 for TV systems).

Third, it consists of using a new transfer function to enable HDR: either hybrid log gamma (HLG) or perceptual quantiser (PQ). HDR usually comes with a fifth improvement: wide colour gamut (WCG). All these parameters are standardised in International Telecommunication Union-Radiocommunication (ITU-R) BT.2100.1. These options add to those available using standard dynamic range (SDR) according to ITU-R BT.7092 and ITU-R BT.2020.3 provides an overview of the different TV combinations of spatial resolution, frame rate, color space, and transfer functions available.

The paper, titled The Agony of Choice, talks about the EBU and IRT organising evaluation of the assorted codecs with more than 200 possible codecs identified.

“After a voting round by broadcasters, more than 40 of these configurations were selected for actual testing. The focus was on real-world products, not laboratory prototypes. Seven generations were encoded for each configuration,” notes the paper.

“The peak signal-to-noise ratio (PSNR) of all sequences was calculated. Subsequently, the first, fourth, and seventh generation of each sequence was scored by expert viewers. The results indicate, that overall, the tested codec configurations perform well up to the seventh generation. For some of the algorithms, a small quality degradation was seen in noise behaviour and spatial resolution but no difference in dynamic range was noticeable. There is a small discernible difference in the performance for the different transfer functions,” adds the advice.


  •   
  •   
  •   

You must be logged in to post a comment Login