SMC Mosaic Repro Now with Version Two

[avandonk]

Registar is a fine program but is limited by being 32 bit so it has memory limitations. It can only stack about eight 190MB tifs.

I tried the latest of version of Deep Sky Stacker and it will stack 20+ 190MB (1.6x native sensor size) tiffs and use when needed all eight (hyperthreading) processors on the Intel i920. I used Images Plus to convert raw to fit and correct for darks and flats. I then converted the unstretched fits to 16 bit tiffs. It is far better to correct for darks in the fit form as no interpolation has occurred so it is far more accurate. This of course only applies to one shot colour.

Deep Sky Stacker then only used to stack these tiffs. Image of screen shot to show the settings. It did a very good job of minimising noise with the Median Kappa-Sigma rejection. This is very cunning as it replaces the pixel values outside the set parameters with the current median value. This is far better than removing noise later by some sort of smoothing as this also loses real signal. The RGB Channels Background Calibration does a very good job of getting very close to correct colour balance.


The advantage of stacking upsized files, especially if they were dithered is that you enhance resolution. In fact it is equivalent to doubling the number of pixels on your sensor.

One problem that this causes with a 300mm optic is that at the corners now the RGB signal of each star does not exactly coincide due to slight residual CA. I just split the image into RGB tiffs and used Registar to realign them. I also used star reduction on the RGB images R 1.4, G 0.0, B 0.9. This removed the colour haloes.


Camera Canon 5DH, Canon 300mm F2.8L at f/5, Hutech LPR filter. Fridge at -10C.

Panel one RHS, Exposures 20x( 1m, 2m, 4m ) and 24x8m at 1600 ISO.
Panel two LHS. Exposures 20x( 1m, 2m, 4m ) and 22x8m at 1600 ISO.

New version large image 10MB.

http://d1355990.i49.quadrahosting.co...1/smcmos_R.jpg

Old version large image 9MB.

http://d1355990.i49.quadrahosting.co...10/sncmosd.jpg


You can see that there are far more well defined faint stars and noise is lower.

I am sure with practice that I may do better.


Bert

[Octane (Humayun)]

Kappa-sigma stacking is all I've ever done. It is a godsend.

In my widefield of the Milky Way, I was imaging it when it was low on the horizon, and, the bottom of the last 5 or 6 frames had trees in them. The algorithm successfully removed the trees and replaced them with sky/stars.

Bert, there are all sorts of weird colour casts in the image, with an overall green cast. These are no doubt due to the disgusting things that HDR software does with saturation. Also, a quick benign stretch of the attached image shows so much more faint data that can be brought out.

Otherwise, mighty resolution and pinpoint stars due to the modifications that you've painstakingly made to your rig.

Regards,
Humayun

[avandonk]

Humayun I am only 18km from the center of Melbourne so light pollution and then the gradients they cause are a constant irritation.The variability of of the sky at my place is mainly caused by the Melbourne eddy where all the pollution just circulates around and around.

Once I get all the methods 'perfected' I will collect very dark sky data.

Thanks for the comments. If I figure out something that is useful I always try to pass it on.

I am basically doing astrophotography to keep as sane as I was last week. Having an active mind and body keeps all the life support systems working.

I am really pleased how I can now exceed the resolution of my sensor.

All these methods are useful to others.

Bert

[avandonk]

Here is a better version. I hit all the different exposure stacks with GradientXterminator before using EasyHDR. The FOV is 6.0 X 5.1 degrees.

Large image 11MB

http://d1355990.i49.quadrahosting.co...1/smcmosR2.jpg


Bert

[strongmanmike (Michael)]

Amazing field Bert, excellent job

We can't all get to dark skies all the time oooor all have observatories huh?.....and then when we do get out there we forget things like...laptops

Mike

[iceman (Mike)]

Nice work Bert. The repro is definitely better.