Printable Version of Topic

Posted by: mmokkp Jul 12 2016, 12:41 AM

Hi everyone

I'm not entirely sure if this is the place I should be posting in, but a lot of people directed me here (I'm new).
Recently I created a new mosaic of Callisto using raw data from voyager 1.
I created my own dark frame for this image, that I believe is a bit better than the nasa ones, although I'm still working on them.

http://i.imgur.com/i30059U.jpg

I hope you like it.
Any comments or thoughts would be very welcome

Posted by: EDG Jul 12 2016, 01:34 AM

It looks good, especially for a first go! Though to be super-nitpicky I do see a few little offsets from the component images on the lower limb of Callisto.

(this was originally posted by the OP at https://www.reddit.com/r/space/comments/4s6kgx/anticipating_first_pictures_from_the_juno_mission/ . I was one of several folks there to ask him to post it on this site!)

Would you like to explain how you made the image here? (there's a lot of comments on that thread to wade through to find that info!) What software did you use, what techniques etc?

Posted by: mmokkp Jul 12 2016, 10:47 AM

So, basically I took the raw data from voyager and created my own dark frame for that flyby. Nasa used a couple missed shots as their dark frame and averaged them to create a single dark frame that was then stretched or shrunk in intensity in order to compensate for different exposure time. I created my dark frame using not only data from missed shots, but I also used partially lit frames, where I removed the lit object and used the rest of the frame. This enabled me to create dark frame for basically each individual exposure. I also created “master darks” from darkening frames from much longer exposures and simulating the shorter ones. For this particular image I used a dark frame from 246 frames, but something like 70 of the frames were almost double-precision (2 averaged frames in one), so the quality of this dark frame is probably equivalent to something like stacking 280 frames. Dark frame was subtracted and resulting file was a 16bit grayscale image

Next I created a mask in Photoshop to remove the dotted grid (don’t know proper term for it) and filled all frames with content-aware fill. While this does produce probably unscientific data, it looks much better than averaging from neighboring pixels, and doesn’t look bad on high contrast areas or borders. I didn’t do the geometry correction using this grid.

Before lens correction I interpolated the data with a modified version of waifux2 scaler, with a photo model. This scaler was originally designed for enlarging anime, but a couple people modified the code a little, added OpenCL acceleration and created a photography model for it. The scaler uses deep convolutional neural networks, and is trained on other data before it can scale. I believe the original project now supports other scaling models as well. It can scale really well, much better than any “dumb” algorithm I ever used. I did this so as to not lose any data during geometry correction and during the rotation of images for mosaic stitching.

The offsets on the bottom are probably due to me not using the grid to create initial geometry correction. If you look closely you can also see 2 hot pixels on the left, that were missed by me as well as two dark spots, one to the right of the crater, second below it, where the geometry grid was moving and my mask couldn’t correct.

Posted by: belleraphon1 Jul 12 2016, 12:02 PM

No nits to pic... just a gorgeous image.
Thanks for sharing.

Posted by: Bjorn Jonsson Jul 13 2016, 07:47 PM

Great mosaic and nice to see high quality stuff from someone completely new here - welcome to UMSF. It's seamless and much better than the old '1979-versions'. The only possible issue is that the color looks a bit uneven and strange to me (but image looks awesome if converted to a grascale version!).

Posted by: MichaelPoole Nov 30 2017, 08:27 PM

Great picture and thanks! But what color data was used? Can you do the same image but with "classical" Voyager color please?

Posted by: mmokkp Jan 12 2019, 09:07 PM

Hey, it's been a while, but back in August I updated the color data for this image after MichaelPoole asked me privately. This is the result: https://imgur.com/a/DQaUlOG

While the alignment is still not perfect, it's much better than the previous version. Also this is using full rgb color instead of just clear+green+UV (that this particular encounter shot), originally used in my image. In the previous image it turned out that the image alignment had a mistake that I ignored. Now the color data was merged into the image based on individual frames, instead of making a color data mosaic and the results are much better.

Posted by: Floyd Jan 13 2019, 02:58 AM

Very nice mmokkp. Thank you for all the effort you put into this mosaic.

Posted by: Ian R Jan 13 2019, 02:23 PM

Utterly gorgeous!

Posted by: TrappistPlanets May 6 2021, 02:11 PM

i know i am very late but i found something in that mosaic that got me thinking



did i find a Callisto plume?






i found it in this part of calllisto




anyways NICE

Posted by: djellison May 6 2021, 04:31 PM

That's 100% a mosaic stitching artifact.