© 2003-2006 Mike Russell, All Rights Reserved

Curve Workflows

A workflow, as discussed here, is simply an accustomed way, from beginning to end, of dealing with your "work", also known as image files. It can be as simple as opening the image, doing auto-levels, and closing the image, or as complex and well developed as Björn-Olof's wonderfully documented workflow described below.

Note that most of these are curving workflows, meaning that they emphasize curving operations that do not use masking, sharpening, or other important image manipulations.

This section will be expanded as more people describe their work flows.


Workflow #1

Nominal Curving Workflow - nothing fancy, just shadow, highlight, and neutral

  1. Open the image and perform any obvious, big moves such as crop and resize.
  2. Set the shadow point, if one exists, otherwise use Curvemeister's threshold feature to tighten up the shadows, without sacrificing detail in the shadows.
  3. Set the highlight, if one exists, otherwise use thresholding to make the highlights as light as possible without losing significant detail.
  4. Experiment with lab mode, and steepen the a and b channels if color saturation appears to need it.
  5. Add a neutral, if one exists.
  6. Select an appropriate curve shape in the Lightness curve of Lab, or the Brightness curve of HSB, to set overall brightness. If in RGB or wgCMYK, use the master curve sparingly to achieve the same effect.
  7. If appropriate, do a second pass in another color space. For example, an initial pass in HSB to adjust saturation may be useful for images with atmospheric effects, that have a mixture of highly saturated and undersaturated objects may benefit from a first pass in HSB.
  8. Sharpen and retouch the image.
  9. Always save to a copy of your original!!! Better yet, copy your originals to CD before you start editing.

Workflow #2

MY CURVING WORKFLOW - Björn-Olof och Fredrik Svanholm

This is a "short" description of the approach and thinking, I have developed, when correcting images with curves. I'm mainly using lab, HSB and HSL color models when adjusting images, so my approach will fit these models (curves).

I start with cranking up saturation, roughly to a level that brings color close to what could be considered "real" or to a level I feel pleasing. I'm fine tuning the saturation and color after I have set black- (Shadow), white- (highlight) point and adjusted contrast in the image. I start with the saturation because I find it easier to correctly adjust the contrast, in the image, when it has its "final" saturation.

In the next step, I search for the correct shadow and highlight point.
This is not necessarily the darkest and whitest pixel within the image, but rather the darkest and brightest point I can find in the scene, within the image, that I decide/want to emphasize/bring forward. In short - My aim is not to reveal all information, hidden within the image. I only focus on the image content that I want a viewer, of the image, to see.

I frequently change the shadow- and highlight point, after my initial curving (contrast adjustment). If I, after my initial curving, find that I have too little contrast to distribute in the image; I first try to either move the black point closer to the white point (increasing the angle of the curve and thereby increasing the contrast) or to bring down the curve in the shadow region (the beginning of the S-curve), moving contrast from the shadow region to the mid tones.

After setting shadow and highlight point, I try to distribute the contrast between them. I start with examining the histogram, at the same time as I move the cursor over the image, to get a better feeling for tone distribution within the scene. Assuming a "normal" scene, with roughly normally distributed tones (histogram), I mark three points on the line in-between the shadow- and highlight points.

One point roughly in the center of the distribution (usually half way between shadow and highlight pint), another point half way between the center of the distribution and the shadow point and finally the third point about half way between the center and the highlight point (a place that correspond roughly to a place under the histogram where the histogram reaches maybe 10-20 % of the histograms maximum height). After placing the three points, I make a rough adjustment moving them up or down, to find a place that makes the image look OK. Moving it up to lighten and decrease contrast in mid tones (point on the shadow side), moving it down to darken and increase contrast in mid tones. To fine tune the curve, I move all the points left and/or right.

I usually end up with an image where the first point (shadow side) is moved down and the third point (the highlight side) is moved up. This moves contrast (where curve becomes flatter) from the shadows and highlight to the mid tones and the center of the tone distribution (where curve becomes steeper). And guess what; there it is - the famous S-shaped curve.

When I started with curving, I tried to mimic all different curves like the lizard-tail or the S-curve, but it felt awkward and I had a hard time getting my images look the way I wanted. I quickly left applying specific curves and started to try and read the image and to curve after it instead, the shape just became a consequence of my curving.

I think standard curves are a good way to learn and understand the principals of "curving", but each image is unique and needs a unique curve. I'm personally not applying standard curves on my images any more, but their shapes turn up frequently as a consequence of my curving.

Most images don't have perfect, normally distributed tone curves (histograms), some have two bumps - OK I normally ad 6 control points. All images do not have smooth histograms, but histograms that have flat plateaus mixed with steep slopes. The trick is then to ad control points at the right places and placing the contrast (the steep part of the curve) where it's most important for the image (and to place the flat parts where it's "not seen") - to create a smooth and continues distribution of both tones and contrast, that the eye finds pleasing.

It's very easy to ad control points (to adjust tone and contrast locally in an image), but there is a drawback. It becomes increasingly difficult to create a good smooth and continues distribution of tones over the entire image.

If I find that I have too many control points (I can't create a smooth and continues tone distribution) or too little contrast to distribute in the image, I usually end up using masks (or/and layers) to adjust separate parts of the image with individual curves. One good way to do this is to use individual curves for the shadows, mid tones and highlights.

After the tones and contrast is redistributed, I set neutral point(s) to make the image neutral, fine tunes saturation and finally I adjust color balance and individual colors, to get an image that my eye find pleasing. Adjusting colors is indeed an art / a skill of its own, an "art form" that I've just started to explore.

I have come to two conclusions; 1) Practice is the only way to learn and master it all and 2) I have to practice a lot.



Table of Contents