Deconvolution improves the clarity in high-signal areas of the image. It should be done in the linear stage of image processing.
Tip: Although the greatest impact with deconvolution is with bright areas with the Luminance and HA frames, other filters can also have some benefit with deconvolution.
Tip: Install the DeLinear script from Herbert Walter Astrophotography (skypixels.at) In the linear processing stage, we often need to create a clone then apply a permanent histogram stretch. The Delinear script does the stretch automatically.
Deconvolution Star Mask Creation
We need to create a mask for our brightest stars in the image or else deconvolution will try to enhance the edges of the stars, creating dark rings. The star mask will help protect those areas. We don’t need the mask to capture every star, only the brightest.
- Noise Threshold: This will be a very low number as the mask is created from the master, unstretched image.
- Scale: Adjust this higher to capture the bigger/brighter stars
- Structure Growth
- Large-Scale: Increase so the size of the bigger stars are bigger than the actual stars as we need to protect the edges
- Small-scale: Increase so the size of the smaller stars are bigger than the actual stars to better protect the edges
- Compensation: Increase to control how large the tiniest stars can get
- Mask Generation
- Smoothness: Depending on the size of the stars, increase the smoothness to avoid square stars
- Aggregate: Bases the mask by summing the values of all wavelet layers
- Binarize: Creates a star mask where the stars are the same intensity across the entire star. Helps avoid dark halos
- Mask Preprocessing
- Midtones: Decrease the level to apply a stretch to the newly created star mask
Apply the star mask process to the unstretched image.
Deconvolution Render PSF Image Script
We need to create a PSF (Point Spread Function) image of our stars. Every optical system is different and might not create perfectly round stars.
PixInsight includes a script to create a PSF image (Scripts – Render – PSFImage) to simplify this process.
To start, use the defaults with the image. If it works correctly, it will generate a PSF image by hitting the Create button.
If it fails, the sensitivity level should be decreased to detect more stars. Avoid going too far as it might take a significantly longer time to process.
Deconvolution Signal Mask Creation
We need to create a mask so we can focus our deconvolution process on the high signal areas of our images. To create the mask, we do the following:
- Duplicate the image
- Apply the DeLinear script – this does a permanent stretch to the cloned image
- Launch Histogram Transformation and move the left slider to the right so the peak of the cloned image is at 0%
This will make a very dark image as only the brightest areas will remain. Apply the mask to the master image.
With the PSF image, mask and star mask, it is time to setup deconvolution.
First, apply the signal mask to the image, making sure it protects the background and not the high signal areas.
Next, launch the Deconvolution process
- Select External PSF and select the generated PSF file
- Algorithm: Leave the algorithm to Regularized Richardson-Lucy.
- Iterations: To start, have iterations between 10-20. More iterations takes longer to process. We will initially do many tests to get the settings correct before upping interactions to higher numbers.
- Target: Make sure the target it Luminance since we are still dealing with monochrome
- Global Dark: This will usually need to be reduced. If this is too low, stars will get dark rings. If it is too high, the process will enhance the noise. Testing will find the right value
- Global Bright: Leave at 0
- Local Deringing: Enable
- Local Support: Select the star mask created for this image
- Local Amount: Typically leave this as-is.
- Wavelet Regularization: Is used to provide noise reduction. The more layers, the larger the pixel size, and the less reduction you want to do
Once you found the right settings, increase Iterations and apply. This will take awhile to run.
If we zoom into the high-signal areas more, we can better see the improvement in clarity
Deconvolution is an improvement, but subtle.
After you run deconvolution the first time with 50-75 iterations, try running it a second time with only 10-20 iterations.
The first round of deconvolution improved the clarity of the image. But when you compare it with the second round, the image becomes even more crisp.
The next step is adjust each image so the brightness is similar with PixInsight Linear Fit. This will make it easier to combine the different filters into a single color image.