Astrophotography by Salvatore Grasso

 

No Decon

20 Iteration Decon

70 Iteration Decon

Proceed to Page 2 of Deconvolution
Deconvolution2.html

This Tutorial describes how to deconvolute, and process deconvoluted data in CCDsharp and Photoshop CS.


CCDSharp is Freeware available from SBIG. It does a decent job of deconvoluting data, and is easily accessible to all amateur astrophotographers.


First start by opening the Raw FIT image you would like to deconvolute in CCDSharp.

CCDSharp is a relatively easy deconvolution software to use. Its almost as easy as click and deconvolute. The basic CCDSharp work flow is to click on two stars that are not fully saturated, then press “GO”

Like i just stated, you must check the stellar profiles before you have the software start its process. Make sure the stars you want to choose are not fully saturated. I usually try to choose stars in the 20,000-30,000 ADU range. once you find two stars in this range you can proceed to the next step.

Stellar Profile Button

Once you have found 2 stars in the appropriate count range you are ready to select them so the software can start its process. from the top palette of tools select the cross shaped button. Then click on the two stars chosen in the previous step and click “GO” from the tools palette.  A window will pop up asking how many iterations you want to use. you will make 2 different layers by deconvoluting the image twice. The first time chose “2” iterations. Then open the RAW FIT file again, and chose “7” iterations. 

Depending on the size of your image this can take a while.

When it is done Deconvoluting, save your images as:

(ObjectName)(20Iterations).FIT

(ObjectName)(70Iterations).FIT

this way the files are easy to keep track of.

Select Star Button

Open up the three images in photoshop.

The image with no deconvolution

with 20 iterations,

and 70 iterations.


Paste the 20 iterations as a layer on top of the raw un-deconvoluted data, then paste the 70 iteration layer on top of the 20. When your done you should see your window looking something like this. It would be smart to name your layers like i did so you can keep track of them.


I like to stretch all three layers exactly the same in the beginning. i may do two or three iterations of curves to each layer. The curves adjustment must be exactly the same to each layer so the all layers will appear the same. This will make blending easier later.

I have skipped showing intermediate steps taken to go from the Raw unprocessed data in the last step, to the stretched data you see here.


Now that you have stretched all three layers, you will really be able to see the noise differences between the layers.

Please take a moment to observe the different levels of noise in the different layers here. as you can see the image with the highest deconvolution strength is the sharpest, but its also the noisiest and contains the most artifacts. This is why we use the different strength deconvolution, and blend them into a single image.

Here is where we are going to start our layering.

We are going to layer in the stars and dimmer details of the galaxy from the 20 iteration data, to the un deconvoluted data.


start by making sure that the 20 iteration layer is highlighted. Then click  Select --> Color Range. When the Color Range window pops up, you will select the option “highlights” from the drop down menu. Click “OK” and it will select the stars, and brighter portions of your image.

Next we will expand the selection. I usually use 4 pixels, however depending on the size of your image you may need to use more or less.

After expanding the pixel radius, you will want to feather the selection. Feather by half the value you used in the pixel expansion. since i used a 4 pixel expand, i will feather by 2 pixels.

Deconvolution

Using Deconvolution is a powerful way to sharpen up the image you have acquired. Astronomical Seeing plays a huge role in how resolved our images can be. Using a deconvolution routine is one way astrophotographers can combat atmospheric blurring.

Deconvolution works best on oversampled data. what this means is that the seeing was worse than the image scale you are working at. For example, I image at a scale of .68 arc seconds per pixel; if the seeing is worse than .68 arc seconds, the data is over-sampled.

Deconvolution is an ideal way to sharpen images, however amplification of noise when using deconvolution is an issue.

Deconvolution routines amplify noise and artifacts in an image. For this reason deconvolution works best on data with a high signal to noise ratio.  Also when many iterations of deconvolution are applied to an image, dark halos appear around the stars. 

As astrophotographers we can not fix these issues, however we can work around them. The way we can do this is by using layers with different deconvolution strengths, and blending them all together. This allows us to take the best aspects of images with different deconvolution strengths and selectively incooperate them into a final image. This technique was developed by Ken Crawford, and is what he describes as the Multi Strength Deconvolution Layer Blend. This tutorial is my modified version of Ken Crawford’s deconvolution blend technique.