New DrizzleIntegration Tool Released

Hi Juan,

There must be somthing i dont understand with flats could you please explain why needing specific flats for meridian fliped images.

Your master flat overcorrects the upper left corners of your light frames. Also you have some light frames that are rotated by 180 degrees due to a meridian flip. These frames require a specific flat and are obviously not being flattened well.

Erik G
 
I enclose a small, highly magnified section of a without- and with- Drizzle using a QSI 540 camera on a Takumar 6x7 150mm lens.  The difference is astounding.  Many thanks Juan for this superb implementation!  Only defaults were used.

 

Attachments

  • Capture.JPG
    Capture.JPG
    66.8 KB · Views: 316
When doing a StarAlignment with Generate drizzle data set and with a Projective Transformation registration model, is the Interpolation method (Auto, Bicubic Spline, ...) have any effect on the drizzle results? I am thinking no (i.e. drizzle does not require this sort of interpolation) but maybe I am wrong.

Thanks,
Mike
 
I would like to inquire regarding this new tool which I have used successfully I might add.
There is seems to be some complication when I transfer the data from one PC to another. Somehow, the drizzle files are looking for the original path where they where first created and even though I define the present path (for example during Drizzle integration) the process fails.
I have to do Star Alignment again in order to create the files again in a new directory in the new PC in order to run Drizzle Integration.
Is that normal?
 
mschuster said:
When doing a StarAlignment with Generate drizzle data set and with a Projective Transformation registration model, is the Interpolation method (Auto, Bicubic Spline, ...) have any effect on the drizzle results? I am thinking no (i.e. drizzle does not require this sort of interpolation) but maybe I am wrong.

As you have guessed drizzle does not require or apply any pixel interpolation, so the pixel interpolation algorithm used for image registration has no direct effect on drizzle. However, you need to generate temporary registered images to integrate them in order to update the .drz files with pixel rejection data. Hence, the pixel interpolation algorithm used for image registration has an indirect, small but non-negligible effect on the final drizzle-integrated image due to rejection. Perhaps?but I have not tested this?using an interpolation less prone to ringing than Lanczos, such as bicubic or even bilinear, would make sense for drizzle.
 
Dimitris Platis said:
There is seems to be some complication when I transfer the data from one PC to another. Somehow, the drizzle files are looking for the original path where they where first created and even though I define the present path (for example during Drizzle integration) the process fails.

On the DrizzleIntegration tool, open the Format Hints section and specify the folder where you have stored your data files in the Input directory field (yes, I know, not the best/most intuitive place to put this parameter...). The input directory that you specify will override the original paths stored in the .drz files. This will allow you to move your drizzle data sets freely across filesystems and machines.
 
Thanks to all who worked on this.  The results are very impressive.  I have one question.  Why  do some images come out of the drizzle integration tool with a rotation, while others do not?  It's making it tricky to stitch together my mosaics.

Regards,

Mike
 
This tool is awesome, and I highly recommend trying it on your images.

I've been using a cowboy trick where I rescaled my reference image to 200%+ before registering, but this is way better. I did a small test, using 104 light frames, and after drizzling, I resampled the drizzled image back to normal size, and compared with the normally integrated image. The SNR increased by drizzling from 278 to 412, equalling more than double the amount of exposure time.

I was easily able to see a mag19.3 z=0.77 QSO that was drowned in noise in the normal integration.

Cs

Morten :)
 
Juan


This tools is amazing !!! Even for 15-20 images only !!!
Main goal for me is not oversample the image (it is already 4k x 4k ) but generate a 8k x 8k image to have better star shape and roundness. Then make star reduction and processing, then shrink to 4k x 4k image. This is great !!!




Something I saw during the process :
I make StarAlignment with files calibrated then get suffix "CAL"
I make ImageIntegration with files "CAL-R" and CAL-R.DRZ files
I make DrizzleIntegration with CAL-R.DRZ files
and I see Drizzle tools uses original "CAL" files and not aligned ones ? I think this is better because no interpolation is added twice ?
Do you confirm this use ?


Many thanks
Philippe

 
Yes. We have to avoid doing interpolation, but we need the transformation matrix from SA to make the pixel coordinates match.
 
Can this be integrated into the integration tool so that we don't have to do two steps (integration and drizzle integration)? Or is there a script to do both steps? Thanks!

Chunyu
 
wangc5 said:
Can this be integrated into the integration tool so that we don't have to do two steps (integration and drizzle integration)? Or is there a script to do both steps? Thanks!

I don't think it makes sense to combine the two operations.  It is best practice to run ImageIntegration multiple times to adjust the rejection parameters.  It doesn't make sense to run DrizzleIntegration until after that's done.

Cheers,
Rick.
 
Along with what Rick says, that would be a clear breach of one of the most important design principles of PixInsight: Modularity.
 
Guys, have just come back to using this again on some images after having mastered MureDenoise.  As you note the images may get a touch nosier, so I tried to apply it and my default setting for my STT 8300M do not work... they produce the checkerboard pattern.  If I had gotten this in the past, I was able to hack the settings to get it to work, but I cannot achieve good MureDenoise.

Can any of the math wizards explain this to me (actually I am less in interested in the explanation as I am is denoising, so a quick hack would also be appreciated)  >:D
 
Jerry,

There is no guarantee that you can achieve a good result with the script on drizzled images.

But try tweaking variance scale. Maybe try a binary search? If 1.0 is bad try 0.5. If 0.5 is bad try 0.25, otherwise try 0.75, and so on, halving the search range at each step.

Thanks,
Mike
 
Thanks, will give it a shot.  This is just part of my routine workflow now and I have added the drizzle into that workflow, so there is a slight train wreck in PI negotiating these two procedures....
 
Hello there,
I have an image that suffer dithering noise!
Look at this example:
https://mega.nz/#!iZZESToS!vfLCSok_yMuJu_T6MhON-MTWMeU74yqg4jMrSjBYGKU

I am not familiar with PixInsight but I always process my images using BatchPreprocessing & ImageIntegration scripts.

I read that I can use DrizleIntegration script directly inside BatchPreprocessing & ImageIntegration, but I am confused because I cannot understand if I need to combine all the c_d_r images + drizzle data or drizzle data only.

The guide in the 1 th page shows as last step the DrizleIntegration script, but the step before shows the ImageIntegration scripts...
If I run the ImageIntegration scripts I have the final image, so how can I set the parameters & combine the drizzle data ?
Anyone can explain me better?

Is this script really able to remove the dithering noise from the image, or is it better to shot again with more luck?!
Thanks, Filippo
 
Back
Top