keeping super-whites

Shooting (1080p50) AVCHD with a Panasonic HDC-HS700 and then converting with ClipWrap to ProRes 422 yields levels up to about 109% according to FCP Waveform Monitor. So whether this is a high dynamic range camera or not, I've got super-whites to deal with. I can clip the high Y' values at 235 which will be ugly. I can linearly shrink the [16, 254] Y' range to the standard [16, 235] range -- FCP Proc Amp does this nicely -- which should be less ugly. I can apply non-linear shrinks. Or I can keep the [16, 254] Y' range and try to make something of it.

My first task is to make a demo DVD. It will be 720p50. What happens to super-whites -- Y' values over 235 -- in mpeg2 or h.264 encoding?

What release schemes best exploit the super-whites?

Dennis Couzin
Berlin, Germany

My opinion is to always color correct and creatively legalize your signal based on your delivery format and color space. As far as video delivery for TV screens is concerned, I always stick to usual broadcast standards- no luma or chroma out of range. This is because super white values were created for you to have more leverage in post production. It is not used for delivery, so do not make anything out of it. Different screens will display extreme values differently. Some LCDs may pixelate on super white values, some may show it hotter. Compressor clips values above 100 IRE.



www.strypesinpost.com

Thanks strypes. Compressor doesn't always clip Y' values above 235 (100 IRE). For example, when sending an 8-bit uncompressed 4:2:2 sequence from FCP to compressor, to become 8-bit uncompressed 4:2:2 again, Compressor will preserve the Y' values up to 254. If Compressor clips at Y'=235 when making mpeg2 or h.264 compressions maybe this is required in those codecs, or maybe not. I'll try to find out.

I fully agree that a general video release should be fully "legalized" to avoid bad surprises. But your example of LCDs that show the over-235 values as hotter is tantalizing. That's what's wanted from the super-white video, isn't it? Maybe it can be a goal for some limited releases. If some LCD displays pixilate on super-white values and other displays show them hotter, the difference might come from the player/decoder or from the graphics card as well as from the display itself.

There's a certain pleasure in making "illegal" video when the means of display can be controlled to a visual advantage. I predict digital video will break from the ethos of standardization which sustained television, due to the experiments of savvy young dabblers, small software companies, and also aggressive big companies (like Panasonic which made the 1080p50 AVCHD camera in spite of this being illegal per AVCHD).

Dennis Couzin
Berlin, Germany

Graphics card? I was talking about LCDs as general TV and broadcast monitor displays. You only get out of range values in Y'CbCr. Computers will clip them hard at 100 IRE, because that point is mapped to absolute white point in RGB.

Compressor doesn't clip when you transcode, from what I remember, only when you encode to certain codecs (eg. Mpeg2).


>There's a certain pleasure in making "illegal" video

I think you need to exchange the word "illegal" with "out of gamut". When we say "illegal values", we are not talking about shooting women not wearing headscarves in Al Jazeera, or shooting an anti establishment social documentary in Singapore, where the purpose of art is to push the social fabric and to raise social awareness, and to promote the individual right to freedom of expression.

Video standards exist for a few very good reasons- so people can watch what you create, and another group of people can realistically develop technology based on the standards. In fact a common complaint in video is having too much standards, or the creation of awkward standards to allow for backward compatibility (eg. dropframe timecode). Let me illustrate the point.

So let's say we want to create a video with extreme yellow values, and not much blue. So Graeme Nattress writes a codec with these channels- C=0-255, M=0-255, Y=0-5000, Y'=0-1000 with a frame rate of 27fps, and it works in Graeme's own color space which we'll call "Graeme2010", a frame size of 48K and a proprietary from of GZIP compression. CMYY'? Of course, it resembles CMYK which is used for print, and is better for creating pure yellows than RGB. So to ease your pleasure of making illegal video, we now have a format that is perfectly illegal, and it does what we want it to- lots of definition in the yellow channel, our own frame rate, and frame size. Naturally, no one can shoot anything to the codec, and no one can watch it. Pure joy? You bet!



www.strypesinpost.com

strypes, to your point that "a common complaint in video is having too much standards", practically speaking, an overabundance of standards means no standards. I think this is where digital video is heading and I think that software can be made agile enough to eek tolerable images out of files without doing it "right".

I understand that your example of CMYY' space is ironic, but a small technical comment might still be of interest. The reason CMYK can create purer yellows than RGB is that RGB requires negative B values for absolute yellow. RGB coding could be modified to allow negative values, but so long as displays are built from physical RGB primaries, the negative values must finally get clipped at zero. Interestingly, "illegal" (full 1,254) Y'CbCr coding achieves absolute yellow with Y'=157; Cb=3; Cr=180. You can also get pretty close to absolute yellow with broadcast-safe Y'CbCr. This is because the standard tranformation equations for Y'CbCr to R'G'B' can yield negative R'G'B'. A jury-rigged gamma transform then gets the negative B from the negative B', but the RGB display kills it.

The problem of weak yellows from RGB primaries (and the need for negative B) was noted by J.C. Maxwell in 1855. It is one of the reasons for basing displays on more than three primaries. Some DLP projectors already do this. So they can interpret Y'=157; Cb=3; Cr=180, standards or no standards. Digital video won't stand still.

Dennis Couzin
Berlin, Germany

I don't fully disagree, only largely. This is because with standards, different softwares and hardware manufacturers can work together to get a picture to work. One of the main advantages of broadcast vs the web, is that broadcast video is closer to establishing standards than web video.



www.strypesinpost.com