After doing some colour correction in the VSE I was wondering how the colour wheels affected the image. I guess I should have done this first.
Anyway I made a linear grad image that filled the screen, and placed the bmp on the timeline. You will notice on the waveform and histogram that the blend is not exactly linear (perhaps a different colour space?) but it is our baseline.
Then I added a number of Adjustment layer effects that change only a specific wheel. In each case I crushed the value to black. The first was the Darks on the left most wheel. Below you can see that the effect from this wheel terminates around 25% of the waveform scale and it affects the ramp back up to white (on the right), so it has a long fall off.
Next I crushed only the mids, but as there is no slope or Q point value, you cannot alter the amount of fall off around it. Therefore all the image is greatly affected as there is dramatic fall off all the way to the edges.
And finally I crushed the Brights as well. The result is the same, a gross overall affect. Where the whole image is darkened but that is to be expected as the toe or dark part of the image is already black. But what if I raise the Brights to 2.00…
Now we can a similar fall off from 25%, but this time 75% of the image is blown out or at the top of the curve. So it has a short fall off, the inverse of the black wheel.
All this means that corrections with these wheels cannot be targeted as they are in the comp nodes, when using the curve node. In other software the transition between these wheels Dark-mid, mid-bright can be modified, but this isn’t available with Strip properties at the moment. And if you try the same modifications using the color balance node the results differ a lot. The change point is more predictable and you can achieve better contrast. But the processing is slower. It would be great to have access to the colour tools from Comp Nodes in the VSE strip properties.
I think the discrepancy with the shape of your waveform / histogram not been ‘linear’ with no CC applied ie diagonal is your source gradient, could try this one:
http://blendervse.files.wordpress.com/2011/06/grad_8bit.png
The other discrepancy with regard to waveform shape between Nodes & VSE is to do with Color Management, looks like VSE linearises whether CM is on or not regardless or at least the waveform scope shows the same as Nodes+CM on where as Nodes with CM off shows differing results in the waveform scope compared to VSE.
It was a quick test I did just setting Lift to 0 in the HSL.
There’s also minor differences in waveform curve near the base in the Nodes with ASC-CDL flicking color management on and off. With Offset on 0 & Power on 2.
Haven’t looked at render images for comparison between various settings though
Haha, the grad is generated from Blender as a texture applied to a plane. Oh well, it said it was linear. Yes Colour Management didn’t seem to have much effect on the waveform as you say (in VSE).
“Other software” here is somewhat a misnomer as there is a reference specification on the handling of color as defined by the ASC known as the Color Decision List. It comes with sample images and a specific set of code samples for implementation as well as proper formatting of XML and EDL text to define the properties.
If you are interested in this, I’d encourage you to download the ASC CDL reference from the ASC.
http://en.wikipedia.org/wiki/ASC_CDL
Erf. “As blendervse already noted…”
The point that I’d make is that the handling of those has a specification and that the controls should ultimately arrive at the specification.
The PDF is quite informative and illuminating. One needs to jump through a minor hoop to get access to the FTP site however.
Sorry for the idiotic looking comment.
Yes I guess we can expect more comprehensive cross over between the CC tool sets (VSE and Nodes) as the VSE gets more love. And certainly a standard protocol like ASC CDL would be welcome, but my quible was really access to a more flexible definition of slope in blacks, mids and whites. A curve editor would probably solve that.
And with the unseamly spectacle of tracking and stabilising going on at GSoC, this web site continues to pimp the venerable old VSE
soon the fanboys will claim Blender as a FCPX killer instead of an AE killer! (Tounge firmly in cheek)