Lightroom: Point editing mode for tone curve control hard to use because of non-local effects

  • 1
  • Problem
  • Updated 4 years ago
  • (Edited)
The "point edit" mode for the tone curve is a much welcomed way to exert finer control on the tone curve than possible through the parametric controls.

However, as currently implemented moving a single point has a very non-local effect which makes the mode very laborious to use.

I find myself doing more work trying to patch up the unwanted global effects on the tone curve with more and more additional points then doing the actual adjustments. I find it hard to imagine that this mode gets used a lot even though in principle it is a lot more flexible than the parametric standard version.

The global effect a move of a single point may have on the whole curve (in particular the dark and bright end regions) makes the combination with the TAT practically impossible. Pick two different shades with the TAT (not too far away from each other regarding tonal shade) and move one up and the other down. The biggest effect on the curve will be on areas of the curve you didn't want to mess with.

One should have a way of just moving a local part of the curve up and down, ideally with a way to control the width of the effect.

To replicate what I'm talking about try this:

1. Start from a straight, unedited curve.
2. Insert a point at ~36.9 (X-axis) and move it up to 50% (Y-axis)
3. Insert a point at ~67.8 and move it down to 50%
4. Insert another point at ~50 (between the first two) and keep moving it up and down from 20% to 80%; watch the shape of the overall curve while doing this.

See how the "highlights" and "blacks" parts of the curve are heavily influenced while you are moving that middle point? Because of this, I find myself putting in more and more points to stop areas I don't want to see affected from changing shape. Try yourself to insert points to stop the local effect you want to achieve from having a global effect. You'll find that your additional "stop the effect"-points will themselves cause a non-local effect that you'll have to stop with further "stop"-points, etc.

In other curve scenarios, the non-local effect of moving a single point can be much worse.

I do realise that the tone curve needs to be smooth (mathematically, there shouldn't be any discontinuities in its first-order, or even second-order, derivation). I therefore understand the rationale of using splines and I also realise that LR is not the only program handling tone curves this way. However, I think the effect of moving one point is too global (i.e., non-local).

Still assuming the scenario above, try to put two "stop the global effect"-points at ~78/78 and 88/88 (in an attempt to have the highlight region unaffected by moving the point at 50). Then go back to step 4, i.e., move the point at 50 up and down. See how the shape of the curve between 88 and 100 still changes with large moves of the point at 50? That effect is completely unnecessary to make the curve smooth. It is just a side effect of using a particular kind of spline for the tone curve.

I suggest that effect of moving one point on non-immediate parts of the curve should die off much quicker as it is currently the case.
Photo of TK

TK

  • 531 Posts
  • 109 Reply Likes
  • frustrated

Posted 7 years ago

  • 1
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 376 Reply Likes
I've been very pleased with the point curve in general (in fact, its the nicest I can remember using - nicely balanced between local effect and smoothing curve), however I'd prefer if the TAT tool worked in "parametric" mode even with point curve selected. - It is pretty unmanageable using TAT tool in point curve mode. I'd rather set the shape using the point curve, then lift or lower "regions" using that TAT - more like in parametric mode - no new points added.
Photo of TK

TK

  • 531 Posts
  • 109 Reply Likes
Rob, thanks for your interest. If you want the TAT to work in parametric mode, I think you should put the curve into that mode. I'd regard it as a step back, if I couldn't use the TAT to change specific tonal values. The real problem is the propensity of the curve to develop big excursions, not that the TAT addresses specific points.
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 376 Reply Likes
As it stands, I do switch back and forth between point curve and parametric - point curve for shape, parametric for TAT. Given how TAT works in point curve mode, its too unwieldy for me to even use. So, it either needs to be reworked, work like in parametric mode, or I need a tutor... But, the point curve's non-TAT operation seems pretty darn good to me. Maybe I'm just too easy to get along with (har-har).
Photo of TK

TK

  • 531 Posts
  • 109 Reply Likes
Rob, I hear you, but I'd prefer a true fix.

Regarding the non-TAT operation: I realise that I can almost arbitrarily shape the curve by using many, many control points. I just think that I have to use too many. Even just three points will introduce a slight bend to the ends of the curve. Bringing that bend back to linear requires a lot of control points. My argument is that it should take less control points to tame down the non-local effect.

My motivation is the following: Broad adjustments usually work well with the parametric controls. Only if I have local changes to make, I resort to the point mode. I would then love the effects to be more local.
Photo of Lee Jay

Lee Jay

  • 990 Posts
  • 132 Reply Likes
"I suggest that effect of moving one point on non-immediate parts of the curve should die off much quicker as it is currently the case. "

This is a bit tricky. If you make the effects too broad, you have the problem you described, if you make them too narrow, you end up with a very non-smooth curve which results in image problems. Worse, the "localness" of the curve should probably be dependent on both the number of points and their spread, which would be very hard to program without unintended side effects.

While I've occasionally had the problem you describe, I've been satisfied with the ability to add additional points to pin the parts of the curve I don't want to move. However, I don't use the TAT for this very much. Perhaps the feature request here is to narrow the focus somehow only when using the TAT?
Photo of TK

TK

  • 531 Posts
  • 109 Reply Likes
What tool are you using to get the funky image effects? If it is LR then you might just confirm my point about the current curve behaviour not being very useful. If I use many control points, I can approach the curve I want and the image doesn't look funky. My argument is that too many points are needed to achieve certain curves.

"A spline is already a piece-wise curve with smooth transitions.": There are many spline variants so we cannot make general statements about the locality of the effect when a control point is moved. My point was: The curve should consist of segments. Each segment is constrained to smoothly transition into neighbouring segments. When a control point is moved within a segment, the segment shape adjusts while keeping the transitions to neighbouring segments smooth. The latter must not be achieved by changing the shape of the neighbouring segments. This is how you can achieve a local effect and a smooth overall curve.
Photo of Lee Jay

Lee Jay

  • 990 Posts
  • 132 Reply Likes
"When a control point is moved within a segment, the segment shape adjusts while keeping the transitions to neighbouring segments smooth."

I don't think that's a good approach. What if the point you want to move is right on the edge between segments? It might move the portion of the curve only to the left of your selected point and if you move that selected point one pixel to the right it may only move the portion of the curve just to the right of your selected point. Plus, since the neighboring segment isn't moving, you're going to create a very steep transition between that close segment and the point you are adjusting.
Photo of TK

TK

  • 531 Posts
  • 109 Reply Likes
"What if the point you want to move is right on the edge between segments?": Then you move the control point that sits on the transition (edge) between the segments.

You are right that steep transitions are a possibility, but they are with the current approach as well. Try to place a control point very near to an existing one and move it up and down. You'll create very steep transitions (mainly where you don't want them).

I could live with special "stop"-control points that sit on the boundaries between segments. I could also live with neighbouring segments being slightly affected. In the latter case, the opposite end of a neighbouring segment should not let through any changes at all though.

It really sounds more complicated than it would be. I just want the effect of a local move to die off after a while so that one doesn't have to chase its effect along the whole curve.

It cannot be anyone's intention to change very distant parts of the tone curve across many control points with a move of single control point.
Photo of Lee Jay

Lee Jay

  • 990 Posts
  • 132 Reply Likes
Yeah...I think what you really want it a higher-order spline that's less under-damped than the current one. The one issue with that change is what to do with existing images with tone curve points attached. If the math changes for those you'd have a pretty massive change in the appearance of the images. So, my guess is that this change would require another process version if it were to be implemented.
Photo of TK

TK

  • 531 Posts
  • 109 Reply Likes
I think the problem with the current curve is that it uses a spline with too high an order. It seems that the order depends on the number of points which means that tangents through control points are affected by every control point.

Using a spline with lower order and/or fixing the tangents that are, e.g., two control points away from the manipulated point could achieve a more local effect.

Regarding backward-compatibility: Yes, a process version could solve it or a tone curve control switch which is set to "global effect" by default for all point sets for which it isn't explicitely set to "local effect".
Photo of Rob Cole

Rob Cole

  • 4831 Posts
  • 376 Reply Likes
TK,

A. I think I see what you mean. If you could freeze one section of the curve, and force whatever spaghetti is necessary for smoothness to the other parts, then it would be easier to work on one portion, get it right, then move on to the next portion... Am I close?

Regarding point B - if you've figured out how to use the TAT tool in any sort of controlled or predictable fashion in point curve mode, I'd like to see a video - maybe even pay you to fly to my studio for a private session... until then, I'd still opt for an option to not add points when using TAT - maybe alt key.

R
Photo of TK

TK

  • 531 Posts
  • 109 Reply Likes
Rob, yes you are very close.

Regarding A: I hope my last response to Lee Jay's comment should clarify what I want. There should be no need to explicitly "freeze" sections of the curve because single control point movements should not have an effect beyond segment boundaries (it is to be determined how segment boundaries are defined -- by the user or the software ).

Regarding B: Why don't you fly me over in any event and I'll try my very best to show you what I do? :) Seriously, I think your idea using a modifier key is a very good one. If used in a "modified" mode the TAT could be constrained to adjust existing control points rather than inserting new ones. Another modifier key could be used to have the TAT influence the values of the parametric control, but I think it might be useful to explore how the TAT could adjust existing control points.
Photo of dan dan

dan dan

  • 11 Posts
  • 3 Reply Likes
Just add tone curve to the brush tool so we can do local tone curve adjustments already!