Mapping out hot pixels

[AndyJ]

Like all digital cameras mine suffers from hot pixels at long exposures: the longer the exposure, the greater the number of hot pixels. I'm wondering how I can 'map' these out using a masking technique perhaps?
Taking an example. If I want to capture a picture of star trails I would typically take an exposure of around 45 mins to an hour. Obviously with a DC that's not possible - too many hot pixels and I'm not sure what effect that would have on the CCD. I could instead, with a great deal of patience and time, take multiple shorter exposures of, say 15 secs. All these exposures would suffer from hot pixels that need to be removed and cloning them out is a real pain in the proverbial.
So I stack each exposure into layers in one image, using Lighten blend mode to merge them together? This would give me a final trails image but will still leave the hot pixels.
I'm thinking that if I took a dark frame (capped exposure) for the same length of time (15 secs), dropped the black leaving just the hot pixels (and thus a mask) I could then apply a difference blend: in theory, the hot pixels on the image layer would be exactly the same as the mask and appear as black. This would work if 1) each same length exposure produced the same hot pixels of the same RGB value 2) the resulting black spots really did merge into the background of the desired picture and I'm not left with the same problem as originally, except now I just have black spots to remove.

My camera only suffers from hot pixels, not stuck pixels.

Any thoughts or ideas? (I also asked the question on the DPreview forum as well so I will cross-post any useful information)

thanks

Andrew

[Richard_Lynch]

please post an example.

[AndyJ]

Here we go

This is a black frame exposure of 30seconds. It's of the poorest quality jpeg that the camera is capable of but for examples sake it doesn't matter: it's the same/worse with better quality.

The hot pixels are readily apparent. in fact, the longer you look the more you see. These hot pixels appear in all long exposure images. The longer the exposure, the more hot pixels. Note that some are brighter than others - the pixels get 'hotter' over time. In this picture, the brightest have been hot for quite a few seconds whilst the dimmest have only just got started!

[AndyJ]

Anyone have any ideas about masking these out? Please!

[Doug Nelson]

It's kind of complicated, but since the pixels will always be in exactly the same place on a full-frame image, why not just run a threshold adjustment on your black frame image, invert, copy and paste over any other photo from the same CCD, and set blending mode to darken? This will force all those hot pixels to black.

[AndyJ]

But isn't this the same problem as before except with black pixels instead of coloured pixels? A better situation if the original image is of the night sky I admit, but if the background isn't black/very dark then the black pixels are going to stand out.

Actually, just discussing it makes me think that it's not actually possible! Whatever I do I cannot 'merge' the hot pixels into an average of the background colour so they effectively disappear. Then again, I can't believe that noone in all this time hasn't worked out how to take long exposures with a DC and fix up the image afterwards...

What about using the powertools to drop the black, then select the remaining coloured pixels, grow the selection by 2-3 pixels and use this as a selection mask on the original image to create a new layer, gaussian blur this layer to average out the differences?

Blimey....

[Leah]

I have been told (though never tried) that what you need to do is put the camera on manual, leave the lens cap on, set the exposure to the same settings you used for taking the photo, then take another photo. This should be black plus hot pixels/noise. then you just subtract the hot pixels/noise from your original photo et voila!

As I said, I've not tried it myself, but I know people who swear by it...