Combining long and short exposures of an object

[Jack Harvey]

Sometime one may find that a long exposure is in general good but parts of the image may be over exposed and burned out.  For instance a 600 sec exposure of a bright galaxy may have good signal of the arms and the star field but the core may be burned out.  So if one adds data from a 60 second exposure the definition of the core becomes apparent.  My question is, if you have two master luminance frames of the same object but different exposure times what is the best method to combine these master frames.  The result would then be used with other masters  (R,G,B) to improve the resolution of the core.  Thanks in advance.

[Philip de Louraille]

What I have read and makes sense to me is to normalize the long-duration frame to the short-one. So if your short-duration is 60-second and your long duration is 600-second, you then divide all pixels of the long-duration frame by 10. Then you can combine both frames (after they themselves are corrected for dark, etc...) In this way you do not emphasize noise.
Doing the reverse, multiplying the short-duration frame by a factor of 10 would boost the random (and other) noise in that frame.
Now, of course, you may also want to mask any area of the long-exposure frame that are over-exposed as an additional step but I have never performed it myself.

[Jack Harvey]

Phillipe  This might work however would you not be reducing the signal in the long exposure by division by 10?

I have continued to investigate this issue and looking at image integration see some options that might work but not sure (Juan et al?).  FIrst in Image integration combine with Average but in normalization use adaptive + scaling and in Weights use Exposure time?  Then there is the option to select a Region of Interest whereby I think you could just define the area of the core of a galaxy???

[georg.viehoever]

Isn't this kind of combination supposed to be done by the HDRComposition script provided with PI?

Georg

[Jack Harvey]

I believe you are correct and I tried it with the images I was working with and it error a couple times with a "out of range"  When I got it to work I found I could not stretch the hdr image.  Use of the white point or gamma did not alter the size of the data spike in the histogram window and after many iterations I still have no visible image?

[Philip de Louraille]

Hi Jack,
yes you will be reducing signal (and noise) of the long-exposure frame. But your resulting frame (after combining the short- and scaled-long- exposures) should consist of "linear" data. Which you will boost later.
What you propose seems (to me) that your input frame will not consist of linear data - and I'll add that I am agnostic on that. If your planned output is for the beauty of the picture then a non-linear input frame is not that critical. If, however you plan on getting some scientific data out, I would think you would want a normalized frame as input.

[Jack Harvey]

There is some discussion I found here also  http://pixinsight.com/forum/index.php?topic=1303.0

[Simon Hicks]

Yes, the HDRComposition script does this. Remember that it converts all the images to 64 bit before dividing any signal levels by 10...or whatever factor it needs to. So with 64 bits to play with there is no rounding errors or increase in noise level. And the script really is intended to work with flat, calibrated, linear images.

However, whenever I try to use it, it reports 'Error - Out of Memory'. Which is a great shame because I think this is potentially a really powerful / useful script.

Cheers
         Simon

[Nocturnal]

I used that script on M42 last year but haven't since. It should do exactly what you're asking for Jack.