WBPP Wish: Interactive subframe selection to reject frames manually

jesco_t

Member
Wish

Have an option to let (a minimalist) SubframeSelector pop up after debayer to interactively go through the frame rejection based on the usual suspects (fwhm, eccentricity, stars etc).

Reason

Before WBPP frame I did frame rejection manually - with Blink for the grossly damaged frames and SubframeSelector for the more intricate cases (like filtering for star count to catch mildly clouded frames and eccentricity or fwhm to improve sharpness of the integration result).

With WBPP I can still do my blinking before, but the frame rejection algorithm during WBPP is not adjustable. And it's really relaxed and has no trouble letting frames with 0.65 eccentricity pass (which may be good for SNR but kind of damages star shapes occasionally). If I had a mono camera, I would be pragmatic and just run SubframeSelector on my input subs but I shoot OSC so I would have to do this on the debayered frames which are temporary files in the middle of the WBPP process. I could stop WBPP, run SS, make notes which files are bad and manually remove them from the WBPP input but that's manual and really tedious for long integrations with 200+ subs. Sadly, those huge are the ones were you can cull your frames best because you don't have to fight SNR so bad and actually have the headroom to throw out offending subs. So I feel there's a capability gap in WBPP that could easily be closed because all the ingredients are already there.

Thanks for considering this.
Jesco
 
Wish

Have an option to let (a minimalist) SubframeSelector pop up after debayer to interactively go through the frame rejection based on the usual suspects (fwhm, eccentricity, stars etc).

Reason

Before WBPP frame I did frame rejection manually - with Blink for the grossly damaged frames and SubframeSelector for the more intricate cases (like filtering for star count to catch mildly clouded frames and eccentricity or fwhm to improve sharpness of the integration result).

With WBPP I can still do my blinking before, but the frame rejection algorithm during WBPP is not adjustable. And it's really relaxed and has no trouble letting frames with 0.65 eccentricity pass (which may be good for SNR but kind of damages star shapes occasionally). If I had a mono camera, I would be pragmatic and just run SubframeSelector on my input subs but I shoot OSC so I would have to do this on the debayered frames which are temporary files in the middle of the WBPP process. I could stop WBPP, run SS, make notes which files are bad and manually remove them from the WBPP input but that's manual and really tedious for long integrations with 200+ subs. Sadly, those huge are the ones were you can cull your frames best because you don't have to fight SNR so bad and actually have the headroom to throw out offending subs. So I feel there's a capability gap in WBPP that could easily be closed because all the ingredients are already there.

Thanks for considering this.
Jesco
Hi @jesco_t,

I feel that we're talking about a false problem here. WBPP combined with the weighting system, is capable of properly weighting good and bad frames such that the high-quality frames contribute more than the low-quality ones. Very bad frames almost do not contribute or get even completely ignored.
My suggestion is to get confident with comparing the results you obtain by manually removing the subframes with respect to how WBPP handles this automatically and see if it's worth the effort. I am confident that you'll realize that the difference between the two is negligible and most probably the manual removal leads to an even worse result.

In case you do this kind of comparison, I am more than happy to hear your outcome!

Robyx
 
Dear Robyx.

Thank you for the feedback. I mostly agree with what you say. WBPP is doing a good job at rating subframes and the effect in the integrated image is minor. With WBPP as customisable as it is nowadays, it just strikes me as odd that the frame rejection is black box that's not user accessible (as far as I understand). I think WBPP could be even better if that area would receive the (optional) possibility to be fine-tuned - not necessarily interactively as I wrote.

I did do a (non-scientific) comparison and I think a mildly increased frame rejection does improve my images. I usually don't pixel-peep that much, but I noticed some real bad frames (eccentricity >0.6, compared to <0.5 as the main distribution) going into the WBPP stack that's why I looked. The difference is small - no argument from me there. The left side is all frames as processed by WBPP and the left one is with the worst 9 offending frames removed (processed identically, just WBPP, ABE, SPCC & STF)

M81 Comparison.JPG


To be fair, though, upgrading to a better mount would be the real solution here. A C8 on a HEQ5 is marginal in all but good conditions, but funds are limited... ;)

Regards,
Jesco
 
It's hard to tell by eye since the frames look quite similar; I cannot appreciate any significant difference.

Depending on what aspect interests you more, you should measure it in the integrated masters (before doing any processing) and objectivize the comparison.
More importantly, remember that the result on a single dataset should never be extended to the general case, disregarding who is better between the two methods.

That said, consider that you have several weighting options, including the weighting formula that you can use to address a specific objective. In this comparison, I assume you're looking for the smallest FWHM, a result that can be achieved by penalizing elongated or big stars by promoting these weights in the custom formula panel. It all depends on the balance between the effort you are willing to put in for the benefit obtained.

WBPP blink selection is most probably something that will never be implemented. At least not as a replacement or a simplification of what Subframes Selector already does. I am more willing to accommodate the possibility of reloading already debayered images into WBPP and continuing the processing starting from the registration step, this way one can perform the calibration, CC, and debayer, then measure and select the frames using SS and then re-open wbpp, add the selected frames and continue from the registration. Something like that, but at the moment, we're busy with other priorities :) let's see for the future!
 
An interactive mode for frame rejection or a parameter where you can increase/decrease the threshold of what consider bad frames would be welcomed. In extreme cases with 1 or 2 very high quality shots, the current rejection mechanism reject 99% of other frames and keep a few good frames. an Interactive mode would bring more flexibility to this.
 
An interactive mode for frame rejection or a parameter where you can increase/decrease the threshold of what consider bad frames would be welcomed. In extreme cases with 1 or 2 very high quality shots, the current rejection mechanism reject 99% of other frames and keep a few good frames. an Interactive mode would bring more flexibility to this.
Do you have a dataset where this can be reproducible?
 
Back
Top