An Initial Word About Workflow
We have now learned quite a few processes. However, the order we have learned about them is good for comprehension and building up from the basics. But it isn't particularly good at giving a feel for how these different processes are used in practice to create a finished image.
A workflow simply spells out the order that these processes are used in order to proceed from your original raw data to a product. An example workflow with just the processes we have already covered would look something like this:
Even at this simple level we already have several choices to make. Do we use ABE or DBE? Do we use ArcsinhStretch or HistogramTransformation? And this is why workflows are dangerous. Too many people treat them as recipes that can't be varied. Instead, they should be focusing on what each process accomplishes and what they want to accomplish with the image as they transform it from the raw stacks to a finished image.
In some cases that might mean mixing things up. For example, you might get through stretching the image and realize that the image would benefit from another round of gradient reduction. You could go back and fix it during the original step, and that often is the right answer. But you could simply do another round of it after the stretching.
You are like a sculptor working on a piece of rock trying to bring out whatever you are trying to carve. Or you are like a cook trying to create a great dish. There can be multiple ways to accomplish the excellent final result.
Here is a more complete workflow including processes that we have not talked about yet. This workflow is just for the luminosity portion of the image. We will cover color later on. Color is its own kettle of fish.
While this is a reasonable workflow, again, it isn't the only possible one. Curves could be done before the stretch. Denoise could be done before the deconvolution. Denoise might not be done at all.
We have now learned quite a few processes. However, the order we have learned about them is good for comprehension and building up from the basics. But it isn't particularly good at giving a feel for how these different processes are used in practice to create a finished image.
A workflow simply spells out the order that these processes are used in order to proceed from your original raw data to a product. An example workflow with just the processes we have already covered would look something like this:
- Dynamic Crop
- AutomaticBackgroundExtractor (ABE) or DynamicBackgroundExtraction (DBE)
- Stretch using either ArcsinhStretch or Histogram Transformation or Masked Stretch (or some combination)
- CurvesTransformation to adjust contrast
Even at this simple level we already have several choices to make. Do we use ABE or DBE? Do we use ArcsinhStretch or HistogramTransformation? And this is why workflows are dangerous. Too many people treat them as recipes that can't be varied. Instead, they should be focusing on what each process accomplishes and what they want to accomplish with the image as they transform it from the raw stacks to a finished image.
In some cases that might mean mixing things up. For example, you might get through stretching the image and realize that the image would benefit from another round of gradient reduction. You could go back and fix it during the original step, and that often is the right answer. But you could simply do another round of it after the stretching.
You are like a sculptor working on a piece of rock trying to bring out whatever you are trying to carve. Or you are like a cook trying to create a great dish. There can be multiple ways to accomplish the excellent final result.
Here is a more complete workflow including processes that we have not talked about yet. This workflow is just for the luminosity portion of the image. We will cover color later on. Color is its own kettle of fish.
- Mure Denoise
- Dynamic Crop
- AutomaticBackgroundExtractor or DynamicBackgroundExtraction
- Deconvolution
- MultiscaleLinearTransform (denoise) or TGVDenoise
- Stretch using ArcsinhStretch or HistogramTransformation or Masked Stretch (or some combination)
- CurvesTransformation for contrast
While this is a reasonable workflow, again, it isn't the only possible one. Curves could be done before the stretch. Denoise could be done before the deconvolution. Denoise might not be done at all.