Barry - a colourspace favour if you have the time

L

lacuna

Well-known member
A little while back Barry Green posted an excellent description of colour, or chroma resolution, and how this plays out in a comparison of 4:2:2 colourspace (DVCPRO) vs. 4:2:0 (DV/HDV).

The images of a diagonal red line on a white background clearly illlustrated the difference. Barry (or anyone else who knows how to achieve it in their NLE), if you get the chance could you run a "real world" image through the different codecs - ideally 720p DVCProHD vs 720p HDV - to show how colour resolution shows up in complex images.

I know, I know, just what frame to show is problematic, but I find it difficult to translate the somewhat abstract diagonal line into a scene with faces and a spectrum of colours. I guess like a lot of DVXusers I'm used to working with DV images, and don't get much opportunity to work with high-end equipment/codecs. When I look at my DV stills, it's hard to get a definite fix on just what I'm missing in terms of colourspace.

Thanks
 
Well, that's a test I'd love to do, but it's lot tougher to do, because you need to have an uncompressed/never-compressed source picture to simulate what the camera CCDs would be receiving. And about the only way to get that is to shoot something on an F950 recording the direct HD-SDI output. And that's not something I have access to.

I wonder if I could simulate it by shooting 35mm film and scanning a print on a decent-quality scanner?

I'm experimenting with using my 7-megapixel digital camera, shooting a full-res picture (which is something like 3500x2300 or something) and downrezzing that. Problem is that it will already have been through JPG compression, so it won't be a truly representative image -- a lot of high-frequency fine detail will already have been discarded, and I don't know what the true color resolution of the source file is.

If the source picture isn't full-res 4:4:4, then no comparison based off it will provide any meaningful results at all.

As far as your statement:
When I look at my DV stills, it's hard to get a definite fix on just what I'm missing in terms of colourspace.
Click to expand...
There are many examples on this site of people complaining about "stairstep" edges on their footage, usually on reds. I of course couldn't find any in the search, but they're all over the site.

If you want to see what color sampling does in real-world pictures, look at Adam Wilt's page at http://adamwilt.com/pix-sampling.html. That'll show some examples of how low color sampling at 4:1:1 or 4:2:0 render a circular LED of color. 4:2:2 will render the picture with twice as much resolution.

If you want a simple test you can do at home, try lighting a scene with regular light, and then shoot it again with all red light. The all-red version will be terribly low resolution/blurry compared to the "regular" lighting, because you're relying solely on the lower chroma resolution.
 
Barry_Green said:
I'm experimenting with using my 7-megapixel digital camera, shooting a full-res picture (which is something like 3500x2300 or something) and downrezzing that. Problem is that it will already have been through JPG compression, so it won't be a truly representative image -- a lot of high-frequency fine detail will already have been discarded, and I don't know what the true color resolution of the source file is.
Click to expand...

If you are worried about JPG compression - you could try shooting in RAW format - that doesnt go thro JPG compression at all.

If you want I can send you Nikon NEF (RAW)s or TIFFs converted from it.
 
I don't have a raw format on my camera (it's just a PowerShot SD500).

If you want to send a raw TIFF that hasn't been compressed in any way, I might be able to work with that, with the understanding that it still won't be truly representative because it'll need to be re-sized down to 1440x1080 or 1280x1080.

I can't generate 720p DVCPRO-HD files yet -- I haven't figured out how to get the Avid codec to work with 720p. In the past I've relied on someone else using their FCP4.5 to create the 720p DVCPRO-HD file. I can do 1080 files though, so I could do a 1080 vs. 1080 comparison.

Even then, that's still not fair, because 1080 HDV has no provision for progressive, so all 1080 footage is pre-filtered down to about 820 lines. Yet testing a raw image will give an unfair advantage to the HDV footage.

To make it more fair, I'd have to de-flicker the HDV footage, to simulate what the camera would do to it.

Hmmm.

And even then -- once we get through with all these hoops we're jumping through -- it still is only vaguely useful info, because the recording format is only one element in the whole imaging chain. The lens, the DSP, the latitude, the chips, the sampling system, the gamma curves, they all play such big roles in determining the final outcome...

I think the question everyone wants answered is: how does HD100 footage look compared to HVX compared to Z1? And that's something that can't be answered by doing codec tests. Yes the codec testing is interesting (to me at least) but it doesn't answer the real question.

So I'll probably run some more tests. But please understand that no matter what tests I (or anyone else) run(s), they won't answer the basic question of: how does this camera compare to that camera? That cannot be answered until we have cameras to compare.
 
Okay, I tried using a digicam picture... I resized a full-size picture down to 1440x1080 and also to 1280x1080. I output the pic as HDV in best case, mid-case, and worst-case scenarios, and I also output it in DVCPRO-HD. I then pulled those files back into Vegas and copied 'em to the clipboard. Then I went into PhotoShop and made a document that has each shot as its own layer.

It's a shot of some random people at a beach, I chose it because it has a multicolored umbrella, a red ice chest, and some other colorful items (you did say you wanted to see chroma resolution examples).

In the photoshop document, you can compare any layer against any other by selectively making layers invisible -- so you can directly compare HDV best-case against HDV worst-case, for example, or you can almost directly compare DVCPRO-HD against one of the HDV shots (I say almost, because they're not exactly the same resolution, and shrinking one or stretching the other would be unfair to it.)

http://www.icexpo.com/JVCnDVX/Beach-composite.PSD

It's about a 20-megabyte photoshop document.

Again, this is pretty much for entertainment value only. The source picture had been JPG compressed to 6 megabytes, and I don't know what the color sampling is like in my digicam. Then again, there's been so much stretching and resizing going on that it'd be hard to draw hard-and-fast conclusions. However, as is the case with every comparison test I've run, it's pretty obvious that the DVCPRO-HD version is quite a bit better looking than the all the HDV variants, and most especially when compared to the "worst case" HDV file (and, for those who haven't read the other threads, the "worst case" is an unlikely scenario, just like the "best case" is. If you only want to compare two pics, you should choose the DVCPRO-HD vs. the "mid case" pic. Comparing DVCPRO-HD against HDV's "best case" is giving an unfair and undeserved advantage to HDV, because it's very, very, very unlikely that you'll get "best case" results in any real-world shooting situation).
 
You must log in or register to reply here.
Back
Top