Codec Developers

This page is intended to give information on how Tvids might be applied when doing codec development, for example when developing or optimising a video encoder as used in a semiconductor, software encoder, set-top box, video camera, multimedia device, broadcast hardware encoder or similar.

Applications

Appropriate uses of Tvids in codec development applications would be in development of:

• semiconductors, particularly as may be used as part of the simulation process prior to tape-out
• hardware or software encoders, transcoders and format converters, using custom hardware, GPUs or general purpose processors, or a combination
• set-top boxes
• video cameras
• other multimedia devices
• hardware devices which take uncompressed video as input

The functions that Tvids could be used for with the above equipment are:

* Quality testing: some possible means of quality evaluation are listed below.

How to generate an SDI / ASI / DVI / HDMI real-time stream

UPDATE: Drastic Technologies now supports direct play-out of all Tvids on HD-SDI using Aja and BlueFish boards - contact TestVid for details (links on YUV viewers page).

Note: a white paper covers this particular subject in detail

When doing codec development, in most cases the Tvids YUV files can be used directly.  However, there may be times when a real-time stream is required on an SDI / ASI / DVI / HDMI link.  In this case, the issue is to get the uncompressed Tvids YUV files from disk onto an SDI / ASI / DVI / HDMI interface via a specialised I/O board.

All video servers, many hardware encoders and a large proportion of other equipment have internal hard disks and Gig-E Ethernet interfaces.  This allows the Tvids to be directly copied over the broadcaster’s Ethernet network onto the hard disk, and play-out from there.

Where it is required to produce an SDI / ASI / DVI / HDMI stream as input to other equipment, this can be done relatively straightforwardly, using:

• a high performance PC / Mac
• with high speed RAID hard disks
• with an appropriate SDI etc. I/O board, e.g. from Aja, BlackMagic Design or Bluefish444
• and software to control moving the video from disk onto the I/O interface

All these steps are covered in some detail in the white paper, ”Real-Time Play-out of YUV Video in a Broadcast Environment”

Quality testing of broadcast video

Note: this is only intended as a very brief introduction and list of some of the possible equipment that might be used for video quality testing / evaluation: this subject is already extensively covered by many books, research papers and occupies many people involved in research and development of new techniques.

(This assumes that the video is technically correct, i.e. that the video quality is not affected by incorrect encoding / break-up / appearance of green macroblocks etc., although of course there can be macroblock effects / blockiness due to insufficient bit-rate or encoder inefficiences.)

Some basic ideas

There are two types of quality testing that can be done:

• quantitative, i.e. taking measurements
• qualitative, i.e. evaluation by people

Many types of measurements can be done, such PSNR (peak signal-to-noise ratio): the hardest part with all these measurements is coming up with a repeatable measurement scheme that correlates well with human visual perception.

The last part of the sentence also helps to explain why visual quality testing is so hard:

• everyone has a different perception of video quality
• perception of video quality varies enormously with content (e.g. humans see detail far less and accept blurriness far more in fast-moving objects)
• many other parameters affect perceived visual quality, such: as viewing environment (looking at a window on a computer screen is very different to viewing a 50” plasma screen at home which is very different to a broadcast viewing suite); what expectations people have of picture quality (e.g. movies versus TV news coverage); and extraneous factors, including such things as audio quality.

Equipment to evaluate video quality

There are many software packages which include straightforward technical measurements such as PSNR: this is a starting point but is well-known to not have good correlation with visual perception in many instances.

Some vendors of visual quality measurement equipment are:

Footnote

In the end, when evaluating video quality the questions may come down to looking at the video with an expert eye and making a judgement:

• are the customers likely to notice any visual quality / complain about it / lose customers because of it
• will customers be sufficiently excited by the video quality of the product you offer when compared to competitors, that they will buy your product over others

Function

Description of uses

Comprehensive stress testing

Provide a comprehensive test set of video which includes worst case and ‘difficult’ sequences, to check that the encoder does not slow down and lose a/v sync with particular video features

Efficiency testing

Test the ability of equipment to efficiently compress / handle a wide variety of video:

• to determine what are the ‘sweet spots’ and the problem areas / video types in efficiency terms
• if required encoding / transmission bit-rates can be met on all types of video, while retaining a desired minimum quality*

Functional testing

Testing an equipment’s ability to input and/or generate not only SD (NTSC/PAL) and ‘normal’ HD such as 720p or 1080i, but also to efficiently and reliably produce:

• the versions required for web and mobile applications (perhaps producing / outputting them concurrently with SD / HD)
• versions which may be required in future, e.g. 1080p60 and 3-D stereoscopic

Comparison with competitor’s equipment

Direct comparisons with similar equipment from competitors (or perhaps earlier models of vendor’s own equipment) , comparing functionality, speed, efficiency and quality* of encoding, for example to determine strong and weak points versus competitors’ products

Evaluation of technical approach

To check out new approaches, bit-rates etc. to determine if encoding with particular parameters provides desired / adequate quality* levels

Regression testing

Check out new versions of software on equipment and compare with previous versions encoding / processing speed, efficiency and quality*.

I.e. answer the question is the latest software ‘upgrade’ really an upgrade or are there particular types of video where there is a problem

Vendor

Equipment

Notes

Omnitek

PQA

PixelMetrix

DVStation

Uses PixelMetrix’ QMM quality measure

Rohde & Schwarz

DVQ; DVQM

Uses Rohde & Schwarz own measurement (developed with others)

- various -

Using Sarnoff JND

‘JND’ = ‘Just-Noticeable Difference’ that is a just-noticeable video difference between the uncompressed video source and encoded compressed video.  Sarnoff Corporation developed a means of evaluating visual quality; other companies now incorporate JND into their equipment

Tektronix

PQA500, PQA300

Researched to correlate well with visual perceptual