Hi All,
Some back ground first
Image captured with CCDSoft from Sbig8300mono camera. Conveted to 32bit fits in PI.
Lights: 6x120sec RGB each channel binned 2x2, 6x300sec L channel unbinned,
Dark Frames 6x120sec binned 2x2 and 10x300sec unbinned
Using the great videos by Harry and following some help I search and found from PI website and forums. I manged to Do all processing in PI.
Initally I did not have Dark frames so and my friends and I noticed we were all getting small one red one grean one blue pixel all over our image. We belived these to be hotpixels offset by Alignment. So we got the my friend whos equipment we used captured and sent me the dark frames. I made the masters in PI using the Tutorial i think by Vincent. I did the Calibration and at the end I found I Still was seein the Hot Pixels. After searching the fourms I realised that with PI ImageIntegration process the optimise function by design only elimated real noise. It seems to not remove Hot pixels. So based on advise I tried CosmeticCorrection tool this along with later use of ACDNR seem to remove the hot pixels. I have attached a small sample from one corner of my RGB combined image to show what I mean, this is just before ACDNR is applied, which seems to remove this.
But I wanted to go back and step through each segment of my processing to ensure I really understood the settings and what they were doing, apply some science to it. This is the really coold thing I like about PI, there are many measurements that can be used to evaluate quality rather than just judge by visual appeal. So it thought I will go back to my master darks and extract a Defect map, using the DefectMap extrack tool by NKV (wow some really helpfull scripts from this man). Using this I got the Bad Pixel map, now I am trying to apply to my orignal Light frames. I tired to apply to both the Calibrated (master dark removed) and non-calibrated Lights but the system returns this errro.
Channel #0
Fixing bad pixels, morphological median: 1%
*** PCL Win32 System Exception: At address 000007FEFD12A49D with exception code C0000005 :
Access violation: invalid memory read operation at address 000000001083954C
But if I chose Gausian/Square it seems to work ok. So for now I will do this.
Question why is this error occuring? Is it the update to 1.7?
Next question which is better, to apply Defectmap to non-calibrated Lights or to the Calibrated lights?
Also I am having some gradients in my Lights, which I am removing post Image Integration, now I have learned that I should probably do a DBE Map and apply to all before Image integration. One anomaly I noted was this.
Using Image Integration tool, with Windzorised Sigma Ciping i set the clips to 5 and 5. I noted from another hepfull thread here that the maximum SNR is sqrt(Number of Frames). Well if you use Scale+Zero Offset this is true, if I do 10/10 high/low respectively I get max SNR, such tat J SNR/Maths.sqrt(number of frames) returns .99%. But above 5 it seems to be above .99. But never reaches the theoretical limit of sqrt(number of frames). But I read that to help with Gradients one alternative method to individually doing a DBE, is to use Equalize Fluxes. But this method is returning SNR > sqrt(number of frames). How is this possible? Is it possible that the Gradient is causing some internal calculation error in Average SNR values?
Also is there somewhere where I can find a complete list of Outputs (in Process Console) for Each Process and Script that explains what the output means? Well at least maybe for the ones that affect SNR. Perhaps in future some graphical representation of this method could be provided.
Juan I know you get some criticisim on PI Interface being not to automated, but in my case I find this absolutely perfect. There are many softwares out there that take the Science out of it and you just push buttons to get a desired visual result. I like PI for the fact that it returns some control over each step to the User. I am all for automation, but I think Informed automation is best.
Regards
Fahim
