Vectorscope techniques

[nonophuran]

Hi everyone,

About vectorscopes (in the context of RawTherapee and colour adjustments)

I haven’t done much homework on vectorscope typical uses (other than what first Da Vinci Resolve results show on YT), and I’m curious to know about the workflow people would go through in the case of rather uncertain white balance, be it in the darks, midtones, or highlights.

I have some challenging inverted colour negatives (at least, to me; read: no gray card, no clear grey references, no similar shot + a hate relationship with just “eyeballing” the hues), so any alternative way of looking at the colour data could be of interest, to me ==> I want to explore the use of this tool for proper WB (dead centre?) on inverted film negatives (through the Film Negative module in RawTherapee).

• for midtones, would you simply exagerate saturation, vibrance, chromaticity … to make any shift stand out on the vectorscope diagram?
• what about lights and darks? what kind of reasoning would apply?
• (EDIT) Is there anything that allows to control the “offset” mentioned in Da Vinci Resolve YT tutorials?

Thanks for your comments.

[afre (Alan)]

I think this warrants a new topic and would benefit from it. @paperdigits

[nonophuran]

Thanks for splitting it into a separate case, I was not too sure, @paperdigits @afre

The title is perfect

[afre (Alan)]

I haven’t used the vectorscope in a long time, nor have I tried RT’s implementation. I don’t think I have anything worthwhile to share at the moment. Hope other people will bite now that it has its own topic.

[jonathanBieler]

One way to get a feel for it is to look at similar pictures that are well balanced. In my experience they typically go through the center (certainly if you have grays in your image) with one lobe toward blue and one toward orange (how much depending on the scene & light). Then I also look at the tint (green/magenta). I tend to prefer to have things more toward magenta than green but it’s quite a fine balance.

I don’t push saturation up but you can change the point visibility in the vectorscope options. For dark colors I move my mouse over them in the image to understand where they are on the vectorscope (the histogram & parade are also useful to check color balance there).

[nonophuran]

Hi @jonathanBieler
Thanks for replying.

@jonathanBieler

One way to get a feel for it is to look at similar pictures that are well balanced. In my experience they typically go through the center (certainly if you have grays in your image) with one lobe toward blue and one toward orange (how much depending on the scene & light). Then I also look at the tint (green/magenta). I tend to prefer to have things more toward magenta than green but it’s quite a fine balance.

same here!

My personal challenge is related to fatigue: my vision gets used to subtle colour shifts, esp. when dealing with film negatives:

• I get the colours in the ballpark, …
• and then what? if I’m not satisfied and I end up spending too much time on it, and I know I become insensitive to what are obvious shifts for a fresh pair of eyes (i.e. my own eyes, the next morning)

typical example that, on paper, should not be too hard

• fuji industrial 400
• mid-autumn midday
• a few clouds on a blue sky, shot taken when sun was hidden
• intention: make it realistic (not neutral), i.e. rather blue but maybe not magenta (I’m not shooting expired slides, here)

fuji_indus400_bmw320i.pp3 - Google Drive, fuji_indus400_bmw320i.NEF - Google Drive

[jonathanBieler]

Your edit it definitively a bit toward magenta (e.g. the road is a bit magenta) :

I would do something more like this, but if you like the colors there’s nothing wrong with it :

I had a bit of a green cast in the shadow in my edit (there’s a bit of it left) and I had to use a RGB curve to remove it.

[nonophuran]

Thanks for suggesting an edit.

• about the road being magenta I see you have a French GUI, so I’ll assume you know how un-clean Paris streets can be, depending on whatever was spilled before (dust, diesel particles, remains of leaves, etc.). I agree it is too magenta in my original edit. But at the same time, I would not trust the road or the pavement or the paint for neutral tones: we’re not in Tōkyō.
• Note: I see you’ve played with the dedicated White Balance module. I know it’s counter-intuitive, but the WB, in the case of inverted negatives, is preferrably dealt with at film negative module level.

@jonathanBieler

I had a bit of a green cast in the shadow in my edit (there’s a bit of it left) and I had to use a RGB curve to remove it.

→ do you mean that, whilst watching the RGB histogram, you merely offset the green slope in the RGB curves module so as to slide the dark greens so they overlap dark reds and blues?
When I posted my previous reply, I tried my luck with RGB curves, too (see attached pp3, named differently this time). Seems to save the day
too bad it does not have an “auto” colour button, yet.
D75_2684.NEF.pp3 (14.2 KB)

[jonathanBieler]

I actually noticed the green blob on the vectorscope (it’s the same image as above but with the RGB curve disabled). I was a bit confuse since there’s no clear greens in the image but mousing over it I could see it’s in the shadow. It’s typically the type of things you don’t see anymore after looking at the image for 30s but it very obvious if you toggle on and off the correction. For this the vectorscope is nice both to spot and correct the issue. With film negatives I try to avoid using the RGB curves but sometimes it’s difficult not to.

For the white balance module I think I was just lazy, I usually use the one of film negative.

[afre (Alan)]

@jonathanBieler

Your edit it definitively a bit toward magenta (e.g. the road is a bit magenta) :

I like the contrast of this one though.

[nonophuran]

With @afre we took the discussion offline, and we thought it could go public.

I shared my “final” version of this particular shot: BMW 320 - old and busted coupé | Nikon FA nikkor AI-s 50mm f… | Flickr

• a word about (my current) processing: once in the ballpark, after I’m mostly satisfied with colours, if there are specks to be removed, I export to tiff (haven’t decided on the best (float or not float) settings, yet) and I switched Apple Photos for that (yeah, nasty, I know, not really open-source friendly). I have to say I find most of its auto settings very useful, too: the magic wand is just an example and it does wonders.

With this thread,
I’m trying to see what principles I can apply when I’m colour-correcting and making use of the RGB parade and the vector scopes. I think I got the basics (the theoretical pink/amber skin tone angle, the neutral tones). I’m now thinking what could be the workflow to catch most things, when it comes to colour casts, be they in the darks or in the lights.
@jonathanBieler that point about the “green blob” in the shadows was a nice catch. Helpful.

And I know there’s a separate angle to address that, related to the DCP and the tables (base table, LUT) associated with it. Honestly, it’s kind of a mess, and sometimes, if I’m not happy with the hues, I just give up with the film neg conversion and start afresh, from a clean “no profile” DCP, which is much less saturated (but also “wrong”, from an available signal-to-colour mapping perspective, I suppose - see discussion around this post Nikon D5600 RAW NEF and DCP Tone Curve Option - #28 by Thanatomanic).

@afre kindly reminded me that:

WB never satisfies * me giggle giggle *

I say computational photography and AI are becoming our friends, in some cases if you haven’t had your hands on an iOS device, a mac or a google pixel phone, then I highly suggest you try its auto WB features, based on skin tones, neutral grays, or “ambiance” (kind of like the new auto WB in RT, with correlation techniques). I have most experience with Apple’s magic wand - I find it works wonders on flat-ish edits.

In my Film Negative module case, there’s one thing I want to stress: colour acquisition depends on several factors, so it’s extremely easy to end up with very broken colours (in HSL terms). Not just “wrong WB” as in “shifted WB” (“hey, I’ve used Tungsten in Daylight”). Broken meaning: dark reds incorrect, light greens over the top, mid blues not right either… Some of it also has to do with the inversion algorithm. Lastly, I don’t even know for certain if a colour checker passport method would be enough to have a closed-loop system and compensate for any shift elegantly.

@afre also mentioned I should try CIECAM02/16 + local editing stuff (possibly in ART). I haven’t had the time to read much about it yet in the rawpedia.

Here’s where I am, now.
I’m sharing (from a kodak gold 200 film scan)

1. a “reference” shot of a car (Spring almost Summer noon, sunny, car + clean license plate) which allows me to get the red and blue ratios about right for the rest of the roll in comparable exposure (i.e. film density) and lighting conditions (i.e. temperature + tint)
   D75_2641.NEF.pp3 - Google Drive, D75_2641.NEF - Google Drive
   Side note: the WB set in the WB module is the one of my light box.

2. a shot with zero grays we’re in a park, in similar lighting conditions but you know the drill. Light goes through green leaves, the grass itself is very green, … to me, this is just hard.
   D75_2656.NEF.pp3 - Google Drive, D75_2656.NEF - Google Drive
   i. the Film Negative initially calculated median WB serves as base as I’m incapable of sampling an adequate patch
   ii. then I find it too blue… and I end up looking at the unexposed border to get (with RGB curves module) a neutral dark grey there that will become black further down the pipeline.
   → not entirely satisfied, I wish I could make it a bit warmer but it’s a fine line between too yellowish green and yucky magenta.

3. same as (2), but even harder (for the medium and thus for processing): 15LV out in the sun, 9 to 11LV in the shadows, so in my experience, it’s not easy to get correct colours both in deep shadows and in the sun light. I’m not attaching a side-car .pp3 file as I’m far from being satisfied with my edit. D75_2659.NEF - Google Drive

4. should be easier, but… can’t get the specular highlights right D75_2660.NEF - Google Drive

Let me know if you find obvious approaches to this kind of shots.

Thanks again.

[afre (Alan)]

These were my actually comments:

I think this could be part of the public conversation. It is the process that matters, so don’t be shy in showing your progress. Honestly, the other person’s entry gives the image another (worse in my opinion) cast. This one is better than your original post’s, though I think the shadows should be made a colour closer to grey scale.

The above was referring to the edit shown in the private message.

The tricky part about colour adjustments is that WB never satisfies. No way around making it complicated. One thing to note about any -scope/-gram is that they show you colour distribution and nothing about what colours should be. Keep that in mind when editing. E.g., skin can have almost every type of hue. We are just used to favouring certain ones due to movies and commercials. Human diversity is amazing!

Have you tried the CIECAM02/16 + local editing stuff yet? Amazing stuff. Still in dev I think. ART is better in terms of GUI implementation, so you may want to give it a spin.

To clarify, my comment about CIECAM02/16 and local editing was for RT not ART. ART has local editing as well but named and implemented differently. I suggest we stick with RT since that is what we are discussing. Feel free to ask questions about ART in its respective forum category.