Improved Results from Small GraXpert Processing Change

 A Small Change to DSO Image Processing 

Until recently my DSO images were processed using Autostakkert!3 for stack averaging, GraXpert for gradient removal and Corel Paintshop Pro 2021 (PSP21) for aesthetic touchups. This posting addresses the latter two apps and how they treat images. 

GraXpert numbered steps are 1. Load Image 2. Crop, 3. Background Extraction (gradient removal) 4. Denoising 5. Saving (Stretched Processed file). This sequence suggests one should click each step with any parameter setting if needed. Having assumed the latest denoising version would be the best and fastest I was surprised that version 3.0.1 could take 5 minutes to denoise one image adding incentive for noise reduction in PSP21. 

Having recently discovered Denoise version 2.00 produces satisfactory noise reduction in 20 to 30s I decided to try denoising in GraXpert. Unfortunately the sequence often resulted in excessively stretched images and histograms when imported into PSP21 if the normal sequence was followed.

A simple change to the GraXpert sequence fixed this issue giving me incentive to reprocess many past DSO images. Instead of using PSP21 for noise reduction the GraXpert process is changed as follows. After step #3, Background Extraction the image is saved to a stretched processed tiff file. Stretch is set to No Stretch and the stretched processed tiff file is reloaded. Then denoising is applied. 

Now, when loaded into PSP21 the processed image shows no excessive stretching and the noise is greatly reduced along with improved signal to noise. Thus more nebulosity can be pulled out of the otherwise noisy background. The three examples below show the advantage of the the modified sequence. 

Loaded image file, stretched 30%,  reduced gradients, denoised, saved stretched processed file, cropped in Paintshop Pro. Result: Added stretching (brightness) with added noise despite GraXpert denoising.

Loaded image file, removed gradients, skipped denoise, saved stretched processed image. Applied default digital noise removal in PSP21, then cropped. Result: Ineffective noise reduction despite application of PSP21 default digital noise removal. 

Loaded image file into GraXpert and stretched 30%. Removed gradients. Saved stretched, processed image. Set stretch to none then reloaded stretched processed file. GraXpert denoised. Saved processed image then loaded it into PSP21 and cropped. Result: Original stretching and greatly reduced noise yielding smooth dark sky. 

Summary: When using GraXpert to process DSO images save the stretched gradient processed image after step 3. Set stretching to None. Reload processed image. Apply denoising; version 2.00 recommended for speed. Save processed image. 

Globular Clusters Galore (07/12/2026)

Celestron 8 AVX, Celestron 0.63x reducer/corrector, ZWO ASI294MC Pro, Orion Starshoot Autoguider (300mm F5 guide scope), Imaging: 15x30s or 30x15s subs depending on target. Processing: Autostakkert!3 stack averaging. GraXpert: histogram stretch, gradient removal, noise reduction. Paintshop Pro 2021: final aesthetic touches.

Messier 4

Messier 9

Messier 10

Messier 12

Messier 14

Messier 19

Messier 62

Messier 69

Messier 70

Messier 80

Messier 107

NGC5897

NGC6293

M81/M82, M42, NGC2023 area (4/11/2026)

 

M82 and M81

M42

NGC2023 area

Imaging: Celestron  8/AVX, Starizona Night Owl 0.4x focal reducer, ZWO    ASI294MC Pro 60x30s subs. Processing: Autostakkert!3 stack averaging, GraXpert gradient and noise removal, Corel Paintshop Pro aesthetic touches. 

Update: Processed using modified GraXpert processing sequence. 

NGC1333, NGC1624 (1/20/2026)

NGC1333

NGC1624

     

Celestron 8/AVX, Starizona Night Owl 0.4x focal reducer, ZWO ASI294MC Pro, Orion Starshoot Autoguider, Imaging: 120x30s. Processing: Stack averaged Autostakkert!3, GraXpert gradient and noise removal, Corel Paintshop Pro aesthetic touches.  

Reprocessed with modified  GraXpert processing sequence

IC348, NGC1579, NGC1624, Sh2-219 (01/19/2026)

IC348

NGC1579

NGC1624

Sharpless 2-219

Celestron 8/AVX, Starizona Night Owl 0.4x focal reducer, ZWO ASI294MC Pro. Imaging: 120x30s subs. Processing: Stack averaged with Autostakkert!3, gradient removal with GraXpert, Final touches: Corel Paintshop Pro 2021. 

10/10/2025 Infrequently Imaged Globulars + Helix Nebula (NGC7293)

The original plan for the night was to image infrequently or never imaged globular clusters that hover low near the southern horizon. Near the zenith 30 to 60s exposures are used where I do imaging.  With skyglow from surrounding towns near the horizon limited exposures to about 15s. Unexpected intrusions from trees to the south left some planned targets for another time. NGC7293 was not part of this night's plan but made for a fun and revealing experiment with the GraXpert app. Usual setup: Celestron 8/AVX, Starizona Night Owl 0.4X Focal Reducer, ZWO ASI294MC Pro. Exposures: 30x15s subframes. Post processing: stack averaging with Autostakkert!4. Stretching and gradient reduction with GraXpert 3.02, Final processing: Paintshop Pro 2021. 

Messier 19

Messier 22

Messier 28

Messier 54

NGC 6712

NGC7293 (Helix Nebula)

A Bit on GraXpert A.I.

The GraXpert app uses sophisticated gradient removal algorithms including Kriging, Splines, RBF and a recently introduced A.I. algorithm. The non-A.I. routines require an editable sampling matrix be applied to the image. For the A.I algorithm you need only load the image, set the stretch value then run it. However, there is a smoothing parameter that can be set to value between 0.0 and 1.0. It's affect on an image can be dramatic and unlike the non-A.I. algorithms there's no sampling matrix or settings to play with to try to improve your image. The above images were all processed with RBF I most frequently use in linear mode. Where needed I tweak the sampling matrix. Below are three images processed with GraXpert A.I. mode using 0.0, 0.5 and 1.0 smoothing settings. I'll stick with RBF or Kriging.

 

0.0 

0.5

1.00

Recovery from problematic experiment.

Recently I experimented with acquiring light frames sans dark frames in SharpCap. The result was ugly.  For reasons still a mystery to this imager Autostakkert! could not correct for vignetting when the session flat frames were applied. 

Autostakkert does offer the option to apply the dark frame during stacking but this did not work either. Hmm? Astro-imagers often promote DeepSkyStacker for stack averaging astro images; so I gave it a try. DeepSkyStacker rejected my .png flat frame as an unknown format  and rejected it's converted to .tif format as the wrong color. These files are readily accepted by Autostakkert! which was used to stack average the images without using the flat.  

Thanks to careful some cropping in GraXpert gradient remove software, the  descent images below were resulted after post processing in PaintShop Pro 2021  

IC5146 (Cocoon Nebula)

NGC7380

NGC7538

NGC7635 (The Bubble Nebula)

Sharpless 112

Setup: Celestron 8 AVX, Starizona NightOwl 0.4X focal reducer, ASI294MC Pro,  Orion Mini Autoguider. 
Imaging: 30x60s subframes, -10C. Processing: Autostakkert! (stack averaging), GraXpert (stretching/gradient removal), Paintshop Pro 2021.