Negative pixel values?
by imaxblue
on
Mar 1, 2007 at 12:14:25 am
I'm having some issues with some cineon files that I import into AE7.0 Pro. The most common issue I've been seeing is negative values in the red channel in highlight areas. For example, a persons white shirt will often be blotched with cyan. When I run the cursor over the effected area and check my info window, I get readings which indicate a negative value in the red channel. How on earth is that possible? It would explain the cyan, but how/why is that happening?
If I switch to 32-bit, the cyan blobs go away, but then of course I can't do anything with them either, as 99.9% of the effects aren't 32-bit compatible. I need to run my cineon projects in 16-bit linear mode.
Beyond the lack of compatible effects in 32-bit, I'm also working with files that are no smaller than 4K - and up to 8K! I can just barely work in 4K at 32-bit so long as what I'm doing is very, very basic. More than that, and I simply run out of memory.
The bottom line is, I need to be able to work with cineon files in 16-bit linear. This negative color value issue is the only problem I've run into so far. Any ideas???
Re: Negative pixel values? by Mylenium on Mar 1, 2007 at 10:42:54 am
Do you get the incorrect values after employing the Cineon converter plugin or before? If you use the converter properly, there should be no problems. The errors would only be "normal" without proper color space conversion as indeed AE cannot make proper sense of the logarithmic color table in Cineon files. The error going away in 32 bpc is normal, as indeed HDR images can have valid negative values., but you wouldn't get technically correct results this way.
[imaxblue]"If I switch to 32-bit, the cyan blobs go away, but then of course I can't do anything with them either, as 99.9% of the effects aren't 32-bit compatible. I need to run my cineon projects in 16-bit linear mode."
I think you are misunderstanding what 32 bpc does and what it is good for. It makes absolutely no sense to have all effects compute in full float and those that are mostly needed in film workflows are 32 bit. So the number of effectively "useless" effects is way below 99,9%. I also fail to see your point on another level: nothing stops you from using the ComPander utility to use 8 bpc and 16 bpc effects with correct results in 32 bpc workflows.
Re: Negative pixel values? by imaxblue on Mar 1, 2007 at 5:28:45 pm
My goal with what I'm trying right now is to degrain and sharpen this shot.
Here's exactly what I've set up:
16bpc project
work space: linear adobe RGB (D65)
linear blending checked on
interp footage "kodak 5218 icc profile
view: proof setup kodak 2383 theatre
proof colors checked on
With this setup, the guy's shirt is a blotchy cyan mess. I selected one particular pixel. Here are it's readings, R:0, G:652, B:1023
If I simply change the project to 32bpc, the blobs go away, and the same pixel now reads, R:-254, G:652, B:2785
I readily admit I don't know exactly what I'm doing here, so I really appreciate any and all help you fine people can provide.
In the past I would simply do all my work in the 16bpc project, and when finished, I'd drop that comp into a new comp. The new comp being my "output" comp. I'd run that new comp through the 'color profile converter', input set at project work space, output selected as DPX Scene - Standard Camera Film, Intent set to Absolute Colormetric. I'd render out my cineon sequence from there. So far that's seemed to work rather well. Until I started getting these issues with certain shots in the bright/highlight areas.
I don't understand how the HDR Compander works. I applied the effect, and the pixel values didn't change at all. If I apply the degrain effect after the compander, I get the same result, it reverts to the 16bpc look of a blotch cyan mess again.
Re: Negative pixel values? by Mylenium on Mar 2, 2007 at 12:27:13 pm
Mmh, there must be something going on with the color profiles. This may be a bug. Once you apply the Compander, you should not get any negative values anymore. They should be clipped at worst. It is supposed to translate all float values correctly to integers and limit the range. If you e.g. have a float value of 1.2 in a HDR image, it does not directly translate to just being full white, because it already is brighter. Without the Compander a non-32-bpc effect might interpret it as too large a number(exceeding 32768) in 16 bpc and either clip values or produce just plain rubbish because it cannot handle those values. I have no ideas further since I too rarley work with film, but perhaps Chris or someone else might have an idea.
