3 Replies Latest reply on Dec 11, 2012 11:00 AM by Mathias Vejerslev

    Hot / dead pixel elimination in later versions of Camera RAW

    Mathias Vejerslev Level 2

      Hi!

       

      I was wondering if we could get some under the hood info on how Camera RAW handles hot / dead pixels. The standard answer 'Camera RAW eliminates these automatically' does not apply I'm afraid. I've seen more and more of these stuck pixels in my final output over the last few CR versions. Now, it could be my camera showing its age, but I'd like to know more about how Camera RAW maps stuck pixels.

       

      I've just re-developed a set of some stage performance from last year, Canon 5D Mark II, high ISO and pushed, yes, but thats my whole niche. There's stuck pixels in all of them in the same spots. So I wondered how camera raw handles the whole mapping / processing of hot pixels. The set of images are here, and there's access to full size images also: http://www.flickr.com/photos/mvejerslev/sets/72157627732199199/ (I did clone out the stuck pixels on just a few of them though). It's a mix of processing version 2010 and 2012. Did not make a difference to the stuck pixels.

       

      Mathias

        • 1. Re: Hot / dead pixel elimination in later versions of Camera RAW
          Noel Carboni Level 7

          I thought I had noticed this too.  I got no solid answer back when I posted on the subject:

           

          http://forums.adobe.com/message/4493772

           

          -Noel

          • 2. Re: Hot / dead pixel elimination in later versions of Camera RAW
            Jeff Schewe Level 5

            Well, the issue with higher ISO images is: what's a hot pixel vs. what's a result of salt & pepper noise–those are different artifacts...

             

            When you start ramping up the ISO (and use PV 2010 or 2012 without luminance noise reduction) the likelyhood of random bright (more or less) pixels (or subgroups of pixels) goes up.

             

            The hot pixel fix is still in LR/ACR...but it becomes a battle between what's a hot pixel vs what's a result of camera noise...and ACR/LR tend to be very conservative when it comes to mapping out a hot pixel (or dead or constantly a certain color pixel).

             

            ACR/LR can really only estimate what constitutes a hot pixel because there's no way of recording the XY position of a single pixel and telling it how to handle the mapping. It's an auto thing and as such it's subject to sometimes not working perfectly...

             

            To really determine the issue, one (or Eric) would need an actual raw file–not something that's already been processed and resample. The more examples of ACR/LR not doing an optimal result the better the likelyhood somebody will look at doing something about it.

            • 3. Re: Hot / dead pixel elimination in later versions of Camera RAW
              Mathias Vejerslev Level 2

              "no way of recording the XY position of a single pixel and telling it how to handle the mapping"

               

              Well I do think theres a probem with the way the detection works if this is indeed the case. Image noise is random whereas hot pixels are very constant, yes? That should make it dead simple to register a stuck pixel, really. If the mechanism needs manual input so be it, but it doesnt work as is (for me), obviously.