Are there Best Known Methods or recommendations for using DBE in regions of high nebulosity or even IFN? Any Do's or Don'ts? Example - Wide-field Orion region - 100 to 300mm.
DBE Settings?
- Default Sample Radius
- Samples Per Row
- Subtraction vs. Division
- ...etc
- Should all areas that have visible nebulosity be avoided?
- Should 'Sample Placement' cover as much non-nebulosity area as possible OR as little as possible?
- ...etc
Thx
I think your statement might tacitly imply that the placement of samples in number (e.g. samples per row) or coverage is the "method" with the recommendations you seek. However, other techniques take advantage of symmetries that exist in the data for which you can make some reasonable inferences. For example if the edges (corners) of your image are attenuated due to vignetting (even after flat fielding) and due to the nebulosity in your image only one corner appears "available." You can apply a sample there and to first order correct the image for vignetting using a circular symmetry.
The follow-up question is usually- but how do I know *that* one corner doesn't have nebulosity where I put my sample?? Well, you don't. At some level your data will have a faint limit of reasonable information (S/N). Chris' solution is to look at other (presumably deeper) images. This shifts the assessment of data from your control to the unknowns of a completely independent measurement from your own. The point is, I think, that if you cannot detect or tell there is signal there- it *is* to first order a reasonable sample for the sky/background. An image doesn't have infinite signal detection. Just as it is customary to choose a black level for the sky (which isn't black)- there is also a value in the S/N parameter space where you choose a limit. The samples you choose make this concrete when you apply DBE.
So sample selection is relative to your own detected things- and the choice of technique deals with the nature of your data. One thing I have found is that identifying any symmetries first (usually circular or linear gradients) and taking care of these- then if necessary with a smaller sample set taking care of what is left... works well. This tends to minimize the total number of samples and makes it easier to choose them in busy/nebulous fields.
The above is all totally an opinion and there are likely even better ways to approach the topic that I don't know about. I also have another experiential opinion that most "gradients" and issues seen in images are flat field errors and not sky gradients- which means the default method for me is "division." But this is an topic for a different thread.
-adam
