Lightroom: Stuck Pixels don't disappear with LR3?

Hi AC,

It would be great to try this with one of your problem files. I have privately sent you instructions which describe how you can send the file to me via ftp.

Thanks,
Ben

I don't believe this changed. Did you view at 1:1 to make sure you were looking at the pixels?

Yes, and they remained exactly at the same place for all the pictures I took that day. Now they don't appear as I pixel-mapped the sensor. But now I can't be sure that my pictures will always be "hot pixel" free .
PS: Of course I shoot in RAW (PEF).

AC, if you're up for testing something, can you try switching to DNG and see if you get the same result? If that doesn't do the trick, an example image would help.

I think the difference is the new Process Version in LR 3 where the new demosaicking has a much finer grain.

It should still be detecting and removing actual hot pixels, though. Could we see an example PEF from before you pixel-mapped the sensor, preferably low ISO. Use www.yousendit.com or www.dropbox.com if you don't have your own website to host a large file from.

I found how. Here are some 100% crops, from print screen. These ones just appeared in this photo (iso 200, 105mm):

This one appeared in all photos that day, before I did the pixel mapping:

@Steve: It just happen with the new camera and the new version of LR. I don't think that pixel mapping is reversible.
@ Lee Jay: My ex-K100D had a constellation of dead and hot pixels (white and green pixels) and LR2 was able to remove them all automatically. This seems not to occur with K-5 and LR3.

Does LR 3.4.1 work ok with your old K100D pixels or do you not have the PEFs from that camera, anymore, either? An example of a PEF from the K100D would be useful to experiment with.

Dead pixels are black -- stuck at 0. Like a missing cornstalk in a cornfield and the corn is of varying height.

Hot pixels are the color or colors of the photosite's involved and stuck at the maximum value for the photosite. Like a single cornstalk in a mowed lawn.

LR detects the hot pixels heuristically (educated guess) on a per-image basis, without the benefit of a pixel map for all images off a sensor like the camera can do.

It is easier to detect a maxed out value of a single photosite when the surrounding photosites will rarely be maxed out in a properly-exposed photo, because there is not really any chance of a single photosite showing its color fully when all the others are much less in a real-life photograph.

It is more dificult to detect a single 0 pixel because photographs typically have dark areas and noise might make some of those 0. And assuming there is read noise, a photosite that is disconnected may still never actually read 0, and instead have random values slightly above 0, so how much harder that would be to classify as stuck-low as opposed to just a dark area of the image.

The older Process Version did some noise-reduction in the demosaicking step, itself, so perhaps dead pixels were removed this way, rather than by the hot-pixel-removal heuristic. That is why I asked if you have a K100D image that has dead pixels that are removed in LR 2 but not LR 3, and/or removed in the old process version in LR 3 but not removed in the new process version in LR 3. Try changing the process version in an old PEF that you know where the stuck-pixels are.

Did the K100D have a pixel-mapping function or is that only with the newer K5? If it is new, perhaps Adobe is trying less hard with a camera that can fix its own bad pixels, because the algorithm may not be 100% foolproof and they'd rather not destroy real image detail if the user can get around the issue another way.

I have a Canon DSLR with a stuck bright-red pixel near the center of the sensor that is visible at ISO 100. There was an early beta of the new process version where this red pixel was visible, but the production version of LR 3.0 onward has removed it, no matter what the noise-reduction settings are.

Hi! That's a good question, I didn't try to develop a K100D PEF file with LR3, yet. I'll give a try when I have some spare time.
I've done some search over the net and other people seemed to have the same kind of problem with LR3, but no mention about the model of camera.
Yesterday I downloaded the trial version of DxO Optics Pro 6.6 in order to see the result with my K-5 PEFs. It's possible to choose whether or not the dead pixels are removed, and it works just fine. In spite of great correction of distortion and aberrations, I prefer LR interface and its color management. It's just a shame that LR3 has this dead pixel (not) removing issue. Can't understand why...