Hi Rodd,
This is a big issue out here. I have been getting ALLOT of advice never to use the STF for permanent stretches. Experienced PI users recommend using curves and manual Histogram manipulations for stretching NEVER THE STF.
Never say never. And I would recommend to be cautious when somebody says never if strong arguments are not provided. Being said this, as we told in the intensive workshop, there are strong reasons to use STF to stretch the image, specially when we follow an LRGB workflow:
- The auto-stretch settings are based on perceptual properties of the image that, at the same time, rely on robust statistics of the image. First, we can set up a sky background level with the
target background parameter. Then, we can choose the aggressiveness of the stretching with the
shadows clipping parameter. These two perceptual factors drive the stretching process of any image,
even if you do it manually without any automation tool. This way, when we have an RGB image and an L image, we can apply a stretching with the same target background and the same aggressiveness to both images. This means that
the histogram stretching parameters will be different for each image, but the perceptual settings of the stretching will be the same for both.- The key problem of the LRGB workflow, and a general rule applicable to other techniques, is that both images should be compatible before trying to assemble the composite. Applying the same auto-stretch parameters for both the RGB and the L images is way more accurate than performing a manual stretching because that compatibility relies on robust statistics of both images. The images stretched with this technique will have the same sky background level and will have a similar noise level, so the LRGB composite will run smoothly.
Apply processes manually does not mean necessarily to apply a more advanced technique. What I propose is a workflow that combines automation -required to reach a precision level difficult to reach manually- with a manual check of the key points. In this workflow, what's really important is to check the stretching that's being applied by STF in the HistogramTransformation tool; you should always check if you're clipping significative data in the shadows. There are images where a small amount of clipping is not important (let's say you clip 0.1% of the pixels) just because they come from noise uncertainty. There are other images where that 0.1% can be very significative, like a small dark nebula that's actually darker than the sky background; in that case you should modify the auto-stretch settings to avoid clipping that nebula. These are details that you should learn from your experience.
On the other hand, this workflow is not more "scientific" at all. I would say it's more artistic, since the art work is partially based on perceptual facts. Any art requires a technique; by making that technique more objective you don't give it a scientific meaning.
Does the STF try to use the whole dynamic range for an image, not leaving any room at the high end for such things as star reduction or stretching tweeks?
Yes, it uses the whole dynamic range, in exactly the same way as you would do with a manual histogram stretching, basically because you never clip the highlights. So that reasoning doesn't make any sense.
Best regards,
Vicent.
PS: Of course, this workflow is not the only good way to stretch your images; there are many other techniques that yield to different results; but this workflow has its unique advantages, properties and goals.
