›

The three new videos on the weighting methods of subs. Are 300 sec subs as good as 600 sec subs?

Roger Smith
edited April 2022 in PixInsight
Hi Adam,
These are 3 great videos, with clear explanations. Many thanks.

It has been argued that more subs of shorter duration are better than fewer subs of longer duration. In Part 3 of the video series PSF SNR weight says they are approximately equal.

After study the graph of Part 3, there might be a good way to determine if 300 Sec subs are really equivalent to 600 Sec subs. Please look at my attached marked up screen print showing how this could be tested using image integration.

Of course after integration there is always the difficulty to confidently or quantitatively determining which is better or if they are equal.  But if it turns out that 25x300 sec subs are equal to 25x600 sec subs, then I can save a lot of imaging time!

I only offer these ideas, and I know you are busy with AIC. So just do what you want with this.

Thanks,
     Roger

Comments

  • BlockHeadBlockHead
    edited April 2022
    Hi Roger,

    This was the entire point of my presentation!

    It is hard...I know it was for me, to understand that weighting by PSFSW does not track with SNR like methods is what is INTENDED. It is correct according to PI developers that SNR methods disagree. So the answer to your question is there IS NO "REALLY" better in terms of weight. 

    This was really hard for me...and I think it will be hard for others to get as well since I think we are used to thinking in a strictly SNR way. But consider if you used SubFrameSelector and you weighted your images by how round the stars were. You can clearly see that 300sec exposures can be weighted more highly than 600sec exposures if the 300sec exposures have rounder stars Right?

    So... studying the graph to arrive at "300 Sec subs are really equivalent to 600sec subs" is going to get you in trouble.

    On a final note- I do believe that the difference you are seeing here might be an issue. I believe that within the PI weighting means there should be a difference on the plots detectable for differing exposure times since not only is the signal changing... but so is the background. 

    So in my video I say this is not the final chapter for PixInsight algorithms...I fully expect more changes.

    You are correct that you can integrate all of he 300sec and 600sec exposures separately and compare the images. You *will* find that that combined set of 600sec will look better in terms of brightness (and probably noise)... but this is only an SNR determination.

    Personally I think this is a difficult place to make a stand. But I think I am accurately describing the PI developer's point of view.

    -the Blockhead


  • Roger Smith
    Adam,
    Great explanation, and no, I do not believe 25 x 300s good subs can be equal to 25 x 600s good subs after integration. I think your conclusion in the 3rd video that each method should be used intelligently depending on your own particular data sets is good advice to us. So if two significantly different subs times in my data, I would not be choosing PSF SNR, but rather NWeight. Now I am pondering if there is any downside to using NWeight for all situations.... Maybe, maybe not.

    Many techniques in PI require usage variations, just as after MureDenoise you can sometimes have some blotchy black areas in the background, but you showed us that cancelling the first layer, then MureDenoise can give a better result. (I hope I said that right). I would not blame MureDenoise. It is just a different situation requiring thinking and problem solving.

    Overall, I come back to a question I asked months ago, is there a definitive (quantitative) way to know if Integration A is slightly better or worse than Integration B. Visually is the current best answer. If you can't see it, well there must be no difference.  

    Roger
    :-B
Sign In or Register to comment.