Z CAM E2 Color correction with plugin

Jason Zhang
4 min readSep 29, 2018

I’ve got a lot of feedback on plugin and people always said they don’t have the habit to use plugin and it’s prefered to use LUT. Here I do think we need to have some detailed introduction to it so you can realize what it can do for you and then you decide whether to use it or not.

We have performed a very quick setup in lab for this topic and we shoot following clips:
1. WDR mode: Normal/+3EV/-3EV

2. Zlog mode: Normal/+3EV/-3EV

I’m using mix light setup so you can see the difference better.

Here is the snapshot of all these original footage for reference:

Let’s do color correction in this way: WB->Adjust brightness by gain->LUT(+3EV/Normal/-3EV)

It’s easy to find out the color and saturation change:

If it’s originally over exposed you reduce gain at log curve it will be washed out.

If it’s originally under exposed you increase gain at log curve it will be over saturated.

This will change skin tone at least and it’s based on very accurate LUT which makes color corrected at highlight and lowlight area. It’s caused by the nature of log curve:

Log curve is like above and there are some characters with it:

  1. In low light area the slope is very steep.
  2. High light area the slope is very flat and it very close to linear.

Gain can simply be explained as multiplying a digital gain to current image. The difference from doing digital gain in camera is in camera it’s all linear. Then if a pixel has a luminance of 10, applying gain 2 means it will be 20. In linear space doing this will not change the color as RGB are the same.

But doing this on curve it will be different. Let’s make a simple calculation on it.

Assume we have two pixels in linear space to be (8,7,8) (R,G,B) and (128,112,128) and then converted to log domain to be (73,64,73) (228,223,228)

Then if we do white balance adjustment in linear space like RAW, we will convert log back to linear space and then apply the white balance to whole image. You will find that no matter we choose the dark pixel or bright pixel the RGain and BGain are the same: 0.9. Then doing WB won’t break the color. Well, this is ideal situation and in real world quantization will have error so choosing too dark pixels isn’t a good idea and 10 bit will be way better than 8 bit.

If we do that in Log space, it will be interesting: If we use dark pixel to calculate white balance, then we need to apply 64/73 = 0.87 but if we choose bright pixel it will be 223/228=0.98. Please remember that this adjustment is for whole image. It means if you make dark pixels white the bright area will be greenish or if you make bright area white the dark/shadow will be magenta in this case.

According to the nature of log curve,doing color correction is the same thing. Increasing gain to make dark area brighter will increase saturation and decreasing gain to reduce brightness of highlight area will decrease saturation. Color will be broken as well. If that’s at color grading phase for art it’s OK but if you are looking to get color corrected that’s a bit difficult.

Let’s see if we make such tweak in plugin what’s the result:

There are slightly contrast difference due to grey scale layout during shot but you can see that color is recovered very well.

We do have LUT to convert log to rec709 and it works well if you just use it to do white balance adjustment. Unless we create several LUT which contains exposure adjustment like what we have in plugin, the over/under exposure compensation will be still at log domain.

Here is the original files and graded version of the test for your reference:
https://drive.google.com/open?id=1tbT8VdYyOlrwUSiVHOx1tu1sBgbGxdbO

--

--