Main >> Hobbies & Interests >> Astronomy

 
CCD Image Processing
More Information Imaging Telescope  Equipment Equatorial Mount Polar Alignment



  
Digital Image
Processing Techniques

PJM's Astronomy Gallery

 

  


 

The objective is not to discuss basic features of image processing, which can be found in many books (see Further Reading) and websites, but rather my findings with my equipment.

Technically, for the best signal to noise (S/N) images, there is no case against routines seeking master dark and flat field images requiring multiple dark and flat field images. Pragmatically, however, this rigorous approach often has to be offset by the considerations:

1. Reasonable to good seeing is fleeting (certainly in the UK) and must be used to take images.
2. The more "target" images the better, since they can be stacked or integrated. The S/N ratio increasing by 1.414 (SQRT[2]) for every doubling of image numbers.
3. With 4 minute or more exposure times, an average of say 5 to 10 dark images, is a long time, when sleep is beckoning.

Consequently, my dark frames are at least one, but more generally an average of two to 3 images. Flat field images are rarely employed for deep-sky objects at prime focus. However, they are usually a must for planet images using eyepiece projection, to remove lens dust particle shadows. In short, if you are entering this hobby, there is no need to be over concerned with image "calibration". Concentrate on optimising the telescope to CCD image scale, good polar alignment and good target tracking.



Whether it is deep-sky, planets or other targets, my intention is to keep image enhancement processes minimal, since they all remove information. A simple linear scaling to change the brightness and contrast is the ultimate. Not surprisingly, I have not found a unique set of procedures to apply to all images. But I do use a common basic procedure, which produces reasonable results, I hope you will agree.

Use Pix-M5 for Basic Processes
1. Subtraction of dark frame and division by flat field
2. Merging of images if not mis-aligned (otherwise Photoshop or lately AIP4win)
3. Subtract background.
4. Non-linear stretch to power 25 or more, and check for sky background illumination from scattered light, clouds etc.
5. Copy and Paste to Photoshop. Image scale from 510 by 290 to 510 by 373 pixels.
6. If satisfactory Save As a JPEG file for website.
7. If sky background illumination threatens image then process as follows

1. This illumination is generally not symmetrical as with vignetting and not readily removed by making a mask layer using the Gradient Tool.
2. Use the Rubber Stamp (Clone) tool to copy the illuminated area across itself, and the target image if it is affected, such that all bright stars are eliminated, Do not change density of illumination.
3. Apply a Gaussian blur to this layer of 1 to 3 pixels, select by trial and error.
4. In Apply Image subtract the blurred layer from the original. Merge visible layers.
5. Save As a JPEG file for website.

The process is illustrated in the images of the Comet WMI LINEAR. Note the Moon illuminated left side of the unprocessed image compared to the processed image. Dark frames are an average of 5 images at 1 and 2 min..
 -
 =
Mask prepared by Rubber Stamp tool and 3 pixel radius Gaussian blur.
Final image after 100% subtraction of mask layer in Apply Image.
Stacked 1,1, 2 and 2 min. images. Uneven sky illumination from the nearby Moon.

Initial processing is again made with Pix-M5 as follows:

1. Linear stretch white range from 256 to 300 if image is near saturation.
2. Unsharp mask large features at pixel radius: power between say 3:3 to 6:6
3. Unsharp mask smaller features at pixel radius between say 2:1 and 1:1
4. Copy and Paste to Photoshop. Adjust Image Size to 510 by 373 pixels.
5. Possible unsharp mask tweak followed by noise filter and brightness adjustment.
6. Save As JPEG file for website.
Unprocessed Saturn image on left. Image on right following Linear Scaling from 256 to 300 and Unsharp Mask at 3:3 in Pix-M5. Subsequent Unsharp Mask in Photoshop at 100% and 1.5 pixels.
 

1. Open separate R, G and B (and L) raw greyscale images in Photoshop.
2. Open New image in Photoshop of same greyscale size but in RGB mode.
3. Copy and Paste the greyscale images to the corresponding Red, Green and Blue Layers in New RGB image.
4. Examine R, G and B histograms. Linear stretch the lower valued images to the median value of the highest median pixel value using Brightness adjustment.
5. Align R, G and B images and examine combined and hence colour  image. Adjust Image Size to 510 by 373 pixels.
6. Adjust colour through Colour Balance to white stars and black to deep blue sky.
7. If desired Paste luminance image and align. Adjust opacity of this layer, as desired.
8. Merge layers and export and Save As GIF file to retain maximum colour information.
I am presently experimenting with AIP4WIN as a processing tool. Though disappointing, as yet, it is perhaps early days to form a judgement.
 

R. Berry and J. Burnell, The Handbook of Astronomical Image Processing, Willmann-Bell Inc.(2001)
M. Mobberley, Astronomical Equipment for Amateurs, Springer-Verlag (1999)
D. Ratledge (editor), The Art and Science of CCD Astronomy, Springer-Verlag (1997)
S. B. Howell, Handbook of CCD Astronomy, Cambridge University Press (2000)