astroimages
Darin E. Fields' Astrophotography Pages
 

Some Thoughts on Image Processing

As one becomes more and more involved in capturing astroimages with any CCD device, the need for and nuances of processing those raw data frames becomes apparent. And when you are working with a noisy device like a Quickcam, the need for data manipulation after capture is even greater. As the novice discovers, good images from a Quickcam start with lengthy sessions at the camera tinkering with the settings of a program like Dave Allmon's QCV2, but end with extensive processing at the desktop.

What I have learned is that my first attempts at image processing--which often yielded less than desirable results--are sometimes not as hopeless as I first thought when revisited. Experiment, experiment experiment--that's the key. Mistakes and botches are sometimes as useful as more positive results. Below I discuss some Photoshop techniques I have been experimenting with recently and getting some decent improvement from older images--that's what I do when its cloudy, reprocess old images (hey, who says re-runs are boring!).

My Standard Processing Routine

My purpose here is not to provide a primer on processing--many are readily available at the QCUIAG website and elswhere on various CCD imager's websites, all much better than what I might provide. My basic techniques have been gleened from these sources and generally run as follows:

Once I have obtained my median dark frame (usually from an equal or greater number of raw dark frames than image frames), I increase the size of that dark frame by 400% (using bicubic resampling) in Photoshop. Raw image frames are also resized by 400%, and then the median dark is substracted from each of the image frames and saved. For a detailed discussion of resizing and the reasons for doing so, visit Antonio Cidadao's excellent super resolution page. I might note that before I start this process, I regularly sift through the raw image frames and discard any that show image blur or excess noise. My goal is to have a set of consistent, clear raw image frames for dark subtraction and stacking.

m11raw: m11dark:
Sample of a raw image frame and a raw dark frame of M11 at 5 seconds. 20 image and 20 darks were taken.

The next step is to align and stack the now dark-subtracted image frames. For that I use Dave Allmon's DarkGen program. The stacked image--still at 400% of the original size--is then ready for more processing in Photoshop.
Sample of an aligned and stacked frame ready for further processing.
m11rawframe:

This raw stacked frame is then subjected to processing in Photoshop. Typically such processing includes histogram stretching and sharpening before being resized at 25%--thus bringing it back to the original frame size. Occasionally I will conduct still more adjustments after resizing.

Final M11 image after processing.

Some Re-processing Experiments

Recently, I have been experimenting with adding some processing steps to this normal procedure. Specifically, at the start of the process I run a sharpening function on the raw, stacked frame, following by a "fade sharpening" function applied as a "screen."

Original alinged and stacked image frame.
m11rawframe:
Result of applying a sharpening function which is then faded as a screen.
m11applyscreen:

The next step involves basic histogram stretching. Normally after I have performed a histogram stretch, I would move right to final tweaking. Instead, I added another step in which I basically applied the image to itself (using the "apply image" function) as a mask with the mode set to screen (instead of the default "normal"). This process yields the following results:

Image after standard tweaking using histogram stretching.
m11histogram1:
This is the result of then applying the image to itself as a mask in screen mode.
Select "apply image," check the box marked match, and set the mode from "normal" to "screen" in the Photoshop apply image dialog box.
m11applyimage:

Additional histogram tweaking and final processing can now be applied. Typically, I sharpen the image, or I run an unsharp mask before resizing by 25% to obtain the final product. So what's the outcome? Below are a side-by-side comparison of the two processes.

m11a:
Final image frame using standard processing procedure.
m11new2:
Final image frame with the additional processing steps added