Making Sense of the Myriad Choices in Production CODECs

With the continued evolution of frame rates, resolutions, color spaces and transfer functions, identifying the optimum choice of CODECs has become progressively more difficult. Their compression technologies are used both for encoding data, compressing it for transmission and storage, and decoding it, decompressing the data for playback and editing. Each of the hundreds of combinations of CODECs and settings offers different results in the balance of quality and complexity, optimizing competing requirements like image quality, bitrate and available processing power. This requires not a single codec per manufacturer, but CODEC families designed for different products, functions and priorities. It poses a real challenge to broadcasters, aggregators and service providers  who may feel inundated with options.

In their article, “Ultrahigh-Definition Production CODECs – the Agony of Choice,” available in the May SMPTE Motion Imaging Journal, media technology engineers from the Institute for Broadcasting Technology (IRT), the European Broadcasting Union (EBU) and the British Broadcasting Corporation (BBC) tackle this challenge. The authors explain how a team from the IRT and EBU designed a testing methodology that incorporates both objective measurements and subjective evaluations derived from expert viewings to help its members choose the best CODECs for their operations.

They tested a selection of 40 CODEC configurations as determined by vote of industry specialists, focusing on real-world products rather than laboratory prototypes. Their multi-generation approach involved cascading the CODEC (repeatedly encoding and decoding to stress the CODEC’s compression algorithm) while applying a pixel shift between generations to simulate processes like editing and color grading. For each CODEC configuration, they encoded/decoded seven generations, corresponding to a real-world file-based environment of Acquisition, Ingest, Editing, Grading, Archive, Re-use/Editing, and Playout Server. The test sequences included different levels of complexity, detail and movement, and a variety of genres such as sports, concerts, documentary, drama and opera.

The main focus of the testing was on picture quality, one of the key attributes determining the overall performance of the compression algorithm. For each CODEC, they tested every possible combination of spatial resolution, frame rate, transfer function and color space, a total of 18 source sequences. The objective measurement was calculated as the Peak Signal-to-Noise Ratio (PSNR) for each of the seven generations. For the subjective measurement, EBU experts (facilitated by a non-voting CODEC manufacturer representative) scored the 1^st, 4^th and 7^th generations for Resolution, Noise, Chroma Resolution, Dynamic Range, Blockiness, De-noising and Average Performance.

While the detailed results and reports are available for EBU members only, the authors of the article share that overall the tested codec configurations performed very well, with a small expected decrease of PSNR over the seven generations. Subjectively, some configurations exhibited a small quality degradation in spatial resolution and noise, but no difference in dynamic range, while other compression algorithms experienced a small discernible difference in the performance of the various transfer functions. In general, from the 4^th to 7^th generations, the perceptible differences between CODECs varies from “nearly identical” to “quite comparable” versus the uncompressed source sequence.

For a deep dive on testing UHD codecs, read the entire article in this month’s SMPTE Motion Imaging Journal.