From: Milton Aupperle <milton@outcastsoft.com>
Date: June 15, 2006 4:45:45 PM MDT
To: Astro_IIDC@yahoogroups.com
Subject: Re: [Astro_IIDC] Image scale (was: Monochrome Unibrain results)
Hi Alan;
On 15-Jun-06, at 3:33 PM, Alan Friedman wrote:
Hi Milton - I am working on a talk that I'm giving later tonight for Niagara Centre RASC, so some of this is fresh in my mind.
On Jun 15, 2006, at 4:47 PM, Milton Aupperle wrote:
Then you probably need to reduce the pixel matching tolerance when using the longer focal length, especially with average / mediocre seeing and high gains.
Also using a larger pixel area does not always work as well as smaller area for matching on features. This is especially true for low contrast images with high gains, because you wind up matching the background gain noise rather than feature of interest. So it's better to use say a 32x32 or 64x64 pixel area and partially contain a "feature" a larger area with flat noisy data. However what constitutes low contrast is debatable.
All good advice. I had set matching tolerance at the lowest level, but at a large image scale, the combination of noise, low contrast detail and image shift/morphing due to the atmosphere works together to cause trouble. You might find this comparison interesting:
http://www.geocities.com/alanfgag/jupiterscalecomp.jpg
these are single frames through an IR pass filter - on the left with 10" aperture at 7 meter fl, on the right, 6" aperture at 3 meter fl. The bottom images are stacks of about 200 such frames process with unsharp mask.
The main issue I see is that the 10" at 7 meters has considerably higher pixel noise levels.
Astro IIDC actually records the average inter frame noise levels for all the frames and prints it out in the stacking log text notes as the "Pixel Noise Percentage" value. Multiple that value times 2.55 and that's how much your pixels are "jumping" between each frame on average. So if it's a value of 5, then your pixels are changing +/- 12.75 in value between subsequent frames. The higher your gains, the higher this value will be.
When Astro IIDC does it's pixel comparison, it finds the two images that match with the lowest average difference, but with an inter frame noise levels is +/- 12.75 (out of 255) and with low contrasts basically anything is going to match up.
At the larger image scale, sharpness varies by region within a single frame and this region moves around a lot from frame to frame. Multiple region alignment (supported in Astro IIDC v3.0) can be used to advantage to improve resolution of the final result. At the smaller image scale, it makes little difference over a single alignment - and the detail in the festoons is contrasty enough to get a very good stack alignment on a 32x32 pixel square placed in the center of the disk.
About the only thing that could be done would be to add in a small 1 or 1.5 pixel radius gaussian blur to reduce the amplified noise and then sharpen the result using say a 3 to 5 pixel unsharpen mask, which would enhance the coarser features and give it something to "lock" onto. This would be done on the image used for pixel matching not the actual stacked images. It would be similar to what we do for the "Sharpen Frames by" and "Noise Reduction" options. However doing that for each frame would slow things down a fair bit (calculating 2 gaussian blurs at 1.5 radius and say 3.0 radius in series and then doing the multipler).
However nothing will compensate for atmospheric turbulence and if the image ripples and jumps then matching simply will never work.
TTYL..
Milton J. Aupperle
President
ASC - Aupperle Services and Contracting
Mac Software (Drivers, Components and Application) Specialist
#1005 - 815 14th Avenue. S.W.
Calgary Alberta Canada T2R0N5
1-(403)-229-9456
www.outcastsoft.com
