Image of Manoj Koushik publicated in Image Processing Challenges - Linear and nonlinear postprocessing of an image taken with Moon and that has gradients - Main notes: CanonBandingReduction for vertical banding, DBE to remove gradients, Morphological Transformation to reduce stars. Date: June 2016.
I used in this processing four previously integrated images provided by Manoj; Red_Driz_Master, Green_Driz_Master, Blue_Driz_Master and Syn_Lum_Master. The filters used for the images are:
Astrodon 31mm Ha 20nm FWHM: 12x900" -20C bin 1x1
Astrodon 31mm Stromgren V: 12x900" -20C bin 1x1
Astrodon 31mm Stromgren Y: 12x900" -20C bin 1x1
The gradient present in the images can be removed using DynamicBackgroundExtraction with the settings detailed in the following screenshots. Although the gradients are different in each image, I have used the same DBE instance which worked perfectly well in all of them. I could also have applied DBE after combining the channels.
Using ChannelCombination tool generate the RGB image
As can be seen in the following screenshot there are some vertical banding on the image. To remove the vertical banding use CanonBandingReduction script, but previously rotate the image 90º as the tool can only be applied for horizontal banding. After reducing the banding rotate the image back again.
Apply BackgroundNeutralization tool. Select a preview that represent the background of the image and read the maximum values of pixel from Statistics
Apply ColorCalibration tool. Update the values of the background reference from the preview previously used and let the rest of values by default.
Generate a PSF to be used in Deconvolution process.
Generate a star mask using Starmask tool to protect the cores of stars during deconvolution. To build this mask first clone the image, copy the parameters of the STF to HistogramTranformation tool by draging the blue triangle to the bottom bar of the HT tool and then apply HT to the image.
Once stretched, smooth the _clone image by removing three layers in MultiscaleLinearTransform tool
Now apply StarMask process on the _clone to generate the star_mask using the following parameters:
Protecting the stars with star_mask1 apply Deconvolution tool to the image. This process will stand out little details on the galaxies and improve the medium and low size stars profile.
Non linear stretch using MaskedStretch tool.
After the stretch is reveled that the size of the stars is much bigger in the blue channel.
I will reduce stars applying MorpholygicalTransformation to each channel but first generate a star_mask using StarMask tool to protect the galaxies and background during this process. Also apply a dilation with MorphologicalTransformation to the mask to increase the area where I will work on the stars.
MorphologicalTransformation to red channel protecting wiht starmask:
MorphologicalTransformation to green channel protecting wiht starmask:
MorphologicalTransformation to blue channel protecting wiht starmask:
Join the channels using ChannelCombination tool
Now I will start processing the Syn_Lum image following the same procedure of removing gradient and correcting the banding.
Rotate the image and apply CanonBandingReduction script and then rotate back the image
Non linear strecth of the image using MaskedStretch tool
I will also perform star reduction to the big stars protecting the images with the previous star_mask and applying MorphologicalTransformation
Apply HDRMultiscaleTransform to compress the dynamic reange in the galaxies
LRGBCombination to add the Syn_Lum to the RGB image. Clone the Syn_Lum_Master image to use as mask protecting the background during the LRGBCombination.
Apply SCNR to remove some green using the Syn_Lum_Master_clone protecting the background.
Invert the mask (Syn_Lum_Master_clone) and apply GammaStretch to darken the background
A little of noise reduction protecting with a new mask. To build the mask use RangeSelection tool and then multiply the rangemask by six using PixelMath tool
Apply TGVDenoise to the image protecting the center of the galaxies to noise reduction in the periphery of the galaxies and the background
Invert the rangemask and apply CurvesTransformation to increase contrast and a little of saturation.