Thanks, this is helpful. There's one more trick you could do to take this one step further. When you compare the before/after at 15:09, you can see that the synthetic flat calibration indeed did a very good job with many of the donut artifacts, but it still slightly under-corrected. Bright donuts are darkened but not quite enough to blend in with background, and dark spots are lightened but also not enough to blend in. We like the correction it made, but we need it to take a bigger step in the same direction. In pixelmath terms, if you call these images "uncal" and "synth_cal", then the "step in the right direction" is computed as "synth_cal - uncal". If you were to add this adjustment to the original image, "uncal + (synth_cal-uncal)", you get the calibrated image, because, after all, A + (B-A) = B. In this case we want a new synthetic flat-calibrated frame with similar corrections, just more so, so we can scale the "B-A" bit: "uncal + 1.1*(synth_cal-uncal)". I'll bet that's enough to make the bright donuts and dark motes fade further into the background. Iterate as needed with the correction factor. If you find that the synth flat calibration over-corrected (i.e. originally light donuts now appear as dark motes) then use a correction factor
Very nice tutorial. Thank you for that. When I do that, I create my synthetic flat and just use Pixelmath with the formula: $T * mean (flat) / flat. Drag the blue triangle on the integrated " bad" picture and get the same result but faster. CS Christian 😀
Trust me... we all find things like this well after we needed it. I can't even count the number of times I chose to delete data because of bad flats. Ugh!
Glad you liked it! It has saved me on more than one occasion where I had yet to take Flats and rotated my image train for a new object framing... Like 2 night ago...
Your content is invaluable. I suggest you open a Patreon if you are inclined to do so, or accept donations via a currency application. This is great information (and I know it's not your blog post), but this video is important. Clear skies PA
Thanks for the very kind words. I am considering this and have had lots of emails about it recently. I'm still trying to find a way to do it that I am comfortable with. I know my wife would be happier knowing all the time I put into video production is offset in some way. LOL :)
Thanks, this is helpful. There's one more trick you could do to take this one step further. When you compare the before/after at 15:09, you can see that the synthetic flat calibration indeed did a very good job with many of the donut artifacts, but it still slightly under-corrected. Bright donuts are darkened but not quite enough to blend in with background, and dark spots are lightened but also not enough to blend in. We like the correction it made, but we need it to take a bigger step in the same direction. In pixelmath terms, if you call these images "uncal" and "synth_cal", then the "step in the right direction" is computed as "synth_cal - uncal". If you were to add this adjustment to the original image, "uncal + (synth_cal-uncal)", you get the calibrated image, because, after all, A + (B-A) = B. In this case we want a new synthetic flat-calibrated frame with similar corrections, just more so, so we can scale the "B-A" bit: "uncal + 1.1*(synth_cal-uncal)". I'll bet that's enough to make the bright donuts and dark motes fade further into the background. Iterate as needed with the correction factor. If you find that the synth flat calibration over-corrected (i.e. originally light donuts now appear as dark motes) then use a correction factor
Very nice tutorial. Thank you for that.
When I do that, I create my synthetic flat and just use Pixelmath with the formula:
$T * mean (flat) / flat.
Drag the blue triangle on the integrated " bad" picture and get the same result but faster.
CS Christian 😀
Thanks! I’ll have to give it a shot.
Wish i had seen this last week before deleting a load of data with missing flats!
Trust me... we all find things like this well after we needed it. I can't even count the number of times I chose to delete data because of bad flats. Ugh!
Awesome tutorial! Thanks!
Glad you liked it! It has saved me on more than one occasion where I had yet to take Flats and rotated my image train for a new object framing... Like 2 night ago...
Great video!
Thanks! Totally underrated video in my opinion! :) LOL. Definitely has saved me more than once.
Your content is invaluable. I suggest you open a Patreon if you are inclined to do so, or accept donations via a currency application. This is great information (and I know it's not your blog post), but this video is important.
Clear skies PA
Thanks for the very kind words. I am considering this and have had lots of emails about it recently. I'm still trying to find a way to do it that I am comfortable with. I know my wife would be happier knowing all the time I put into video production is offset in some way. LOL :)