Peter Ward
09-08-2023, 11:24 PM
Just putting the finishing touches to one of the most extensive image processing sessions I have done in the last decade or two.
Armed with 600 odd images from the Exmouth eclipse I flat fielded then stacked around 32, seven frame bracketed exposures (224 images) using
Pixinsight’s FFT image registration routine for the lunar surface, and photoshop layers for the HDR corona.
I found Sean Walker’s graduated opacity image stack in Photoshop worked best for the corona HDR combination, but the high pass filtering no so much.
Tools like Topaz AI tended to bludgeon the data to death, or cause bizarre artifacts, with dithering the best way to reduce noise and preserve fine structures.
The coronal details compare very well to Miloslav Druckmüller’s remarkable processing of the same eclipse (http://www.zam.fme.vutbr.cz/~druck/Eclipse/Ecl2023a/WL_1000mm/0-info.htm).
I suspect my raw data is better (larger unobstructed aperture and higher resolution camera)
, but his "unobtanium" proprietary software is white man magic IMHO.
I employed a new (to me) tool called Noise Adaptive Equalization. This revealed many barely visible features of the corona and lunar surface.
I also balanced the brightness of the scene to better reflect the visual appearance of the insanely bright inner corona and chromosphere
(the latter was visible through Baader solar film!) which caused some magenta flaring even with the sublime AP155 objective.
I chose not to do any “deep” (i.e. 4-5 second exposures) during the brief totality, hence the data gets a bit thin at about 3 solar radii.
That said, very pleased with the re-worked result here (http://www.atscope.com.au/BRO/gallery614.html) (the image has been updated from my earlier post) which very much eclipses my earlier work :) (pardon the pun)
Armed with 600 odd images from the Exmouth eclipse I flat fielded then stacked around 32, seven frame bracketed exposures (224 images) using
Pixinsight’s FFT image registration routine for the lunar surface, and photoshop layers for the HDR corona.
I found Sean Walker’s graduated opacity image stack in Photoshop worked best for the corona HDR combination, but the high pass filtering no so much.
Tools like Topaz AI tended to bludgeon the data to death, or cause bizarre artifacts, with dithering the best way to reduce noise and preserve fine structures.
The coronal details compare very well to Miloslav Druckmüller’s remarkable processing of the same eclipse (http://www.zam.fme.vutbr.cz/~druck/Eclipse/Ecl2023a/WL_1000mm/0-info.htm).
I suspect my raw data is better (larger unobstructed aperture and higher resolution camera)
, but his "unobtanium" proprietary software is white man magic IMHO.
I employed a new (to me) tool called Noise Adaptive Equalization. This revealed many barely visible features of the corona and lunar surface.
I also balanced the brightness of the scene to better reflect the visual appearance of the insanely bright inner corona and chromosphere
(the latter was visible through Baader solar film!) which caused some magenta flaring even with the sublime AP155 objective.
I chose not to do any “deep” (i.e. 4-5 second exposures) during the brief totality, hence the data gets a bit thin at about 3 solar radii.
That said, very pleased with the re-worked result here (http://www.atscope.com.au/BRO/gallery614.html) (the image has been updated from my earlier post) which very much eclipses my earlier work :) (pardon the pun)