Linux image preprocessing project - Asterism

[geoland]

Minor updates

[geoland]

Manual update

[geoland]

More user manual updates

[geoland]

Updates bug fixes and improvements

[geoland]

In this version, significant changes, with the following caveat - I am working with cooled DSLR data.

EDIT: tested with Nikon 3300 DSLR RAW data - no cooling - with observable improvements in the stretched image. 

The manual has been overhauled and should be easier to follow. 

The defect map process has been redesigned

Noise reduction may be applied to the final image

The intent has been to find authentic methods of noise reduction which may be applied non-destructively during image preprocessing

I have tested the new processes over and again. The original binarized defect map is still the default. The better (IMHO), mean-shift method, is activated by setting a 'Noise - radius' - '7' is adequate and a 'Noise - amount' - 0.1% at most. 

1 or more dark frames may be used to create the defect map; which is really a master dark with considerable noise reduction, leaving the white and black pixels intact. A median dark is a fair substitute.

With mean-shift noise reduction of the final image, combined with polynomial stacking, the statistical handling of data appears authentic and visually non-destructive - pixels retain their shape and there is no evidence of smearing, unless the 'Noise - amount' operator is overdone - 0.1 - 0.25% is adequate.

[geoland]

Update: A number of issues associated with increased complexity in the previous two versions are resolved.

 

Asterism_0.0.5.3 is tested current and stable. Upgrade to the latest version is recommended, as it includes a number of improvements;

 

Flat auto correction - this is proving very robust. The default setting of 5 appears to be ideal. The old manual flat illumination correction is retained, for the time being.

 

Defect map - the process seems to work very well, with no issues observed - use a ‘Noise - radius’ of 7 and a ‘Noise - amount’ of 0.1. It may be necessary to alter settings for different cameras, although, tested on 3 Canon model RAW data and Nikon, I see no variation, except clean flat division.

Linear noise reduction - this is the noise reduction routine used by Defect map, applied to the stacked linear image.

 

While Asterism is designed to reduce noise in a number of ways, I have found linear noise reduction very effective. The ‘Noise - amount’ value should not be too high - 0.1

 

Correctly applied, the stacked image processed in StarTools, with Tracking ‘off’, the default noise reduction is adequate.

A ‘Noise - radius’ size, 0 (off) to 30 is available, with sensible radii between 5 - 15; and ‘Noise - amount’ 0.1. Increase the radius for noisy images.

 

The User Manual requires an update, however, the differences are noted in the version available for download.

[geoland]

It's been almost 12 months since the first iteration of Asterism and though it will probably never compete with other packages out there, it does a pretty good job and has been a great learning experience. 

I am fairly happy with the latest version, in terms of usability and features. 0.0.5.5 is tested and stable and includes a significant user manual rewrite. This version has a redesigned 'Processing and options' tab with segregated fields which are easier to navigate. Options layout is better arranged and should be more intuitive, in the order; input and output, main process selection, optional processes, followed by system defaults.

To all those people who patiently put up with continual development changes, mistakes, errors and blunders, thank you. This latest version has been tested thoroughly... with added complexity came deprecation of some of the legacy processes, which as far as I can determine have been fixed. I did think to get away from the super-bias terminology, but it seems to have stuck; and in principle, most know what a super-bias is... the method used for now is described below and uses the same DENOISE script.

Quote

DESCRIPTION: DENOISE reduces the noise in an image. It uses the formula:
# input + gain*(input - mean), where mean is a local mean computed in the 
# neighborhood of each pixel and gain=max(0,(std-nstd)/std), where std is the 
# local standard deviation computed in the neighborhood of each pixel. This is 
# an implementation of the Lee Filter. See Lee, J.S., 1981. Speckle Analysis 
# and Smoothing of Synthetic Aperture Radar Images. Computer Graphics and Image
# Processing, Vol. 17:24-32. source

 

[geoland]

New Asterism_beta has been tested and is stable run from the Asterism folder in a terminal. Largely cosmetic changes. PMStack may be launched from the Asterism window.

When complete, all processes return the user to the Asterism main window, automatically, to perform other tasks. Asterism may be closed with the 'Quit' button. Otherwise in accordance with the user manual, which needs an update.

[geoland]

Latest Asterism stable version with manual rewrite.

[geoland]

Update - new stable version

[geoland]

New Asterism homepage

[geoland]

Now on GitHub