›

Suggestion for videos

Frederic Drevon
in PixInsight
I really enjoy these videos but the files you use are generally of very good quality with low noise frames. For us mere mortals that have limited imaging time (typically 4 to 8 hours per target) dealing with noise without making a miss mess is not easy. Any help on that topic would be great.

Comments

  • BlockHeadBlockHead
    Hi Frederic,

    I am not certain I entirely agree (although I am accused of it often!). For example the NGC 6946 and NGC 6888 are both short data sets and the latter in particular is data from a gentleman that took the data from Detroit. 

    The combination of MURE Denoise and later TGVDenoise goes an awful long way. Using MLT in between to remove the 1st layer is one more fairly aggressive stop. Beyond that some of the heavy lifting needs to come from the simple averaging statistics of having enough frames to combine (roughly 15-20 frames). So although you mentioned 4-8 hours... but it is good idea to have enough subexposures in each filter to get good rejection and such. 

    I have seen that the DSLR/OSC type of images can be low signal noisy messes. This is an area I haven't dealt with as much as the regular CCD type of work. 

    Can you post a link to a picture that shows that kind of noise you are dealing with? You would want to include information about how many frames in each filter, whether you dithered or not, the camera you used (plate scale as well)... etc etc. 

    -the Blockhead
  • Frederic Drevon
    This is data from SH2-104. Taken with a FLI16200 camera. Dither on every frame. 20 frames of each Oiii, Sii and Ha and drizzle integrated (so cannot use MureDenoise). Plate scale is 2.10" per pixel. Frames were 5mn and bortle is 4/5.

    Link to Ha :

    I seem to end up with botches of noise in the background. And it gets much worse with DBE.
  • Frederic Drevon
    I should add.This was the outcome after treatment 



  • BlockHeadBlockHead
    Ah... well a number of things.First you certainly cannot compare any of the data I have been working with to these!! Narrowband stuff is extremely low signal stuff and not of the type of data I have been dealing with... *not* because it is superior data..but because it literally has ten times the signal due to the filter!

    So..some questions...

    1. Why Drizzle integrate here? Do you feel you are gaining resolution? 20 images is only near the LOW end of the number of frames you really want. Drizzle integration is for resolution...but narrowband extended fuzzy stuff? This doesn't make sense to me.
    2. Given the above, why not perform Image Integration without the drizzle upsampling... just regular. Then using MURE Denoise..and then later TGVDenoise (and  maybe MLT if necessary). Your data looks quite good for this and MURE Denoise will get you off to a very good start!! If you then still want your big 2x 9000+ pixel image... simply upsample at the end. I do not think you are getting a resolution boost- there is little benefit to Drizzling an image like this. A small planetary nebula maybe for NB.
    3. Regardless of my opinion, I am not certain how you ended up with the mottle. The noise pattern of your original data in the drizzle integrated image still looks pretty good. I wonder what steps you did to give that mottle? DBE might make it easier to see I guess. I do suspect that Drizzle does give you noise pattern that is more difficult to deal with because of the "holes" that are left due to a lower end number of frames.

    So for the moment...I guess I am blaming it all on the drizzle. 

    -the Blockhead
  • Frederic Drevon
    Thanks. I will certainly try that. I usually use drizzle with a higher number of frames mainly because it provides for some really nice round stars considering my imaging scale. Without drizzle they are just not as nice. In any case, I will see if noise handling gets improved without drizzle. Thanks for the feedback.
  • BlockHeadBlockHead
    When you are ready to upsample the image... use a bi-cubic interpolation... you will certainly get round stars. 
    -the Blockhead
Sign In or Register to comment.