I made a test, how image resolution gets effected by a stacking algorithm. When images are registered, they are moved and rotated by a fraction of the pixel accuracy. Depending on used algorithm, there will be some level of blurring in a final stacked image. In this test, CCDStack software is used to calibrate, register and stack the raw-images. Two similar processing are made from the same material, one with normal workflow and the second with images scaled up 200%.
This is a 100% animated crop from images, stacked from a 100% size and 200% size calibrated frames.
10 x 1200s H-alpha with a QHY9 camera and the Tokina AT-X 300mm camera lens at f2.8.
"Mitchell" algorithm is used for up scaling all of the 200% sized frames.
An animation from a single, cropped and 400% up scaled, 1200s H-alpha exposure.
One image is registered and second not. HWFM in none registered image is 1,6 and in registered 2,4.
Method of register is most commonly used Bicubic B-spline.
Specially, if image are undersampled, like in this example, the blurring effects gets stronger and there is a risk to loose some of the finer details.
Down side of up scaling images to 200% large, is the needed amount of computer power and memory! Images will be four times large in file size. In this case a single frame, saved as a 16bit TIFF, will be about 130 MB.
CCDStack will use much more memory per image since images are internally processed as a 32bit floating point image space. All post processing in PhotoShop is done to a 200% sized version.
I have reprocessed Sh2-119 images with a new method, original versions can be found here: http://astroanarchy.blogspot.com/2011/10/sharpless-119-sh2-119.html
Sharpless 119
In constellation Cygnus, Ra 21h 18m Dec +44 00'
Image is in Natural color palette from the emission of ionized elements,
R=Hydrogen + Sulfur, G=Oxygen and B=Oxygen + Hydrogen.
HST-palette from an emission of ionized elements, R=Sulfur, G=Hydrogen and B=Oxygen.
Technical details:
Processing work flow:
Image acquisition, MaxiDL v5.07.
Stacked and calibrated in CCDStack2.
Deconvolution with a CCDStack2 Positive Constraint, 33 iterations.
Levels, curves and color combine in PS CS3.
Optics, Tokina AT-X 300mm camera lens at f2.8
Camera, QHY9
Guiding, Meade LX200 GPS 12" and a Lodestar guider
Image Scale, 3,5 arcseconds/pixel
Exposures H-alpha 10x1200s, binned 1x1O-III 3x1200s, binned 3x3
S-II 3x1200s, binned 2x2
Total exposure time ~5h
0 comments:
Post a Comment