Color (or colour) correction in Kdenlive - how to use Reference Chart and Vectorscope (and other tools eg. colorchecker)

http://userbase.kde.org/Kdenlive/Manual/View_Menu/Vectorscope

And the links to Granjow's blog on that page might help somewhat.

But maybe you are the man to figure out how to do it and write the manual !!!

If shooting to a gamma encoded 8bit video or image format then white balance really needs to be 'right' in camera so the color adjust whatnot or any white card or 18% grey card and adjust WB in camera. Then shoot. Its too late really adjusting non decently white balanced 8bit video or images after especially with low precision 8bit adjustments like kdenlive has currently, when MOVIT coded filters in MLT is exposed in kdenlive then more precision but still on 8bit gamna encoded source.

If shooting raw image ie: 12bit unprocessed then no white balance or gamma has been baked in so shooting a test frame with color adjust or grey card in the scene helps set WB on raw data in the raw development process.

Interesting read is Shane Hurlbuts suggestions about WB on his blog.

http://www.hurlbutvisuals.com/blog/2013/02/using-color-temp-hd-video/

And also Kevin Wang on Flickr with his Canon picture styles strongly suggesting achieving the look you're after by adjusting color temp in combination with certain picture styles.

http://www.flickr.com/photos/aeronation/collections/72157625101919874/
If you need to reproduce the color of a logo or some memory color then WB before shooting will probably get you closer to that color with minimal grading.

RGB is the primary color space from which all others are derived, both in theoretical colorimetry and in practical hardware (sensors have RGB filters).

601 and 709 are based on different RGB primaries, and the different luma/chroma coefficients are a just consequence of that.

Because the difference between 601 and 709 is mainly in the green primary, the luma coeffs are affected the most.

So, if your chart is 709 and your camera 601, green (and cyan and yellow) will be a bit off. If I remember right, the differnce is mostly in the saturation of the green primary, so its angle may not be significantly off.

As your link says, the angles are important, because the white/grey value is insensitive to rotations, it will only show shifts.

> P.S. ColorRight: hahaha
By laughing, I did not mean to say it is an inferior tool, I just laughed at the marketing, which is trying to sell a "new" idea.

1. RGB Parade, not luma waveform, two different color models.

2a. Undock it, freely resize it and move to a second monitor.

2b. ?

3. IRE is an analog concept. 100 IRE is equivilent to 235 YUV (analog), digital value 235 YCC and 255 in RGB. Waveforms vary, some show upto 100% on the graticule as 100 IRE others as 100% RGB or 100% YCC. Kdenlives waveform appears to showing RGB values and slightly incorrect in the low end, a value of 16 in a test file I use is appearing as 28 on the waveform, so the YCC to RGB calculation seems wrong. A YCC Luma waveform should show upto 120%. http://www.kenstone.net/fcp_homepage/video_levels_nattress.html

4. SOP/SAT effect.

5.?

Try "crop, scale and position" to zoom in on the chart.
However, I find it's UI (parameters) a bit clumsy, the scaling works from the edge (not the center) and shifts are called "tilt", and for big enlargements the shifts are not enough to bring any part in the center etc.

I tried pan&zoom too, but the scopes somehow don't change?? ***

For new shots, try to zoom in on the chart already with the camera.

Having the chart fill the frame, it is easier to identify the parts of the chart on the scopes.

Concerning point [5] above, it seems that the vectorscope, which the author used, has a "zoom in" option, this would be equivalent to chroma gain, or increased saturation.
On the Kdenlive vectorscope, you can do this using the slider on the right side.

*** playing with it some more, I found that the scopes (RGB parade and vectorscope) often do not react to the applied effects (like RGB adjust or SOPSAT), although I can see the colors change in the project monitor.
Switching to clip monitor and back seems to help...

This was with a script compile from 20130206
Compiled a new one today (20130225) Now the scopes are more responsive, but they seem to show the same signal in all three channels, R=G=B in the parade, and the vectorscope always shows a symmetric "mercedes star".

The channels on the scopes seem OK on the 20130206 version, but as I wrote above, they aren't always updated after the addition of an effect.

Disclaimer: I am compiling this on an old, heavily patched Opensuse system... although I changed nothing on the system between the two compilings above.

hi, the edge crop effect under crop & transform works fine for zooming in on the chart. You can control left, right, top & bottom, tick the use project resolution and you get zoom with those images.

Then the 3 point balance effect will let you sample, white, black and 18% grey for primary correction.

For manual correction, as in the first link (what good is....):

The signal in a channel is a product of incident light and object reflectivity. So, if the cast is the result of unbalanced light, the right thing for correcting this would be to change channel gains (the diagonal elements in the linear color transform matrix).

To do this you can either use RGB adjust, and set the mode to "multiply", or use the slope parameters in SOPSAT.

NOTE: the above is strictly true in a linear system. But a multiplication by a constant is still a multiplication with a (different) constant after applying a pure power curve, like gamma - so the above is true in a (purely) gamma corrected system too.
Now, the 709 "gamma" curve is not strictly a power curve, it has a linear section near black, to avoid excessive noise amplification. But the difference is probably small enough to not cause problems here.
Video cameras might use additional non-power curves to manage dynamic range, in that case you might be forced to adjust lows mids and highs separately, or apply appropriate curves.

P.S. apropos the first link, "what good is a macbeth color checker"...
The author works for a competing firm, so obviously he has to conclude that it's worthless :-)
Admittedly, the colors on the checker do not correspond to the markings on a vectorscope, making them less useful when using one.
But still, since the colors are specified (the link contains a link to a pdf file with the numbers), you could use them for a check, or to achieve consistency across shots. In Kdenlive, you can use the Pr0be (under misc) to read out the values. Note that in the table, the RGB values are for sRGB, which, if I remember right, corresponds to rec 601 primaries (not sure about gamma). So if you are using 709, you should transform the table.