PDA

View Full Version : Hot Pixels - How to test for?


driggett
02-04-2005, 06:18
With all this talk of hot pixels I was wondering how you test for them?
Thanks,
Chris

Tom Conte
02-04-2005, 06:41
Take an image of, for example, a dark colored wall. Try a medium-slow shutter speed. Open the file in an image processing program or viewer. Look for white dots. If you see something at maybe 1/15th or faster, you have a hot pixel. If not, you're ok.

Some say put on your lens cap, set to B and take a long exposre. IMHO, most CCD sensors will show hot pixels in that situation and it is not something to panic about. But it is unusual and not good if the same 'white dots' show up on medium to fast shutter speeds.

Tom

Jim Watts
02-04-2005, 07:05
Unless the cameras firmware is programmed to map them out and interpolate over them, you should be able to see them with the Gain (ASA setting) turned all the way up, shoot the back of the lenscap or some severe underexposure, and view the image in Raw mode. "Hot" Pixels will stand out like a little light, and dead pixels will be little black spots.

Used to work in the old days.
Brian,
If the ability to map out and interpolate over hot pixels is purely a firmware issue (I hope so) hopefully we will see a future firmware up grade from Epson to add this mapping to the R-D1 for those of us that have this problem.
I don't find it a major issue on prints as I have 4 only barely visable on these in the even tones at 400iso, and you do have to look hard, but would be concerned if there was any substantial increase. Do sensors get worst in this respect with age? I would not want to spend ages checking and cloning out.

Jim

jlw
02-04-2005, 14:10
For Photoshop users: After shooting some neutral frames, you can use Photoshop's image statistics readouts to get an idea of how non-uniform the image might be.

Note that you need to try several different ISO settings, since pixels that are OK at low settings may go "hot" at higher ones.

Once you've got some neutral black frames, you can adjust them with Levels, save them as channels, and then load the channels for semi-automatic dead pixel retouching. If your results are uniform enough, you can even automate this via an Action.

driggett
02-04-2005, 15:08
Okay ran the test. I only had erf files so converted it into jpeg. The shot was teaken of the inside of the cap at 1600 iso and 1/30 of a sec at f/2. The program says I have 3 hot pixels. I will do more tests at different iso and speeds. SHould I be concered with three hot pixels. They are not by each other.

The report is:
Type X Y Luminance
Hot 2066 295 127
Hot 1368 663 76
Hot 1107 1556 78

Thanks,
Chris

driggett
02-04-2005, 15:19
Reran the test and converted the erf fill to tiff after ready about jpeg compression causing false positives in the test. Here is the results
Type: X: Y: Luminance:
Hot 2066 295 99
Hot 1368 663 64

Thanks,
Chris

pradeep1
02-04-2005, 17:39
Reran the test and converted the erf fill to tiff after ready about jpeg compression causing false positives in the test. Here is the results
Type: X: Y: Luminance:
Hot 2066 295 99
Hot 1368 663 64

Thanks,
Chris

So you have two hot pixels. Not bad. I wouldn't worry about it. As for the question regarding CCDs getting worse over time, there may be a case for that, since I've seen it on some older cameras. Mostly will become a "stuck" pixel, instead of a hot pixel.

Jim Watts
02-05-2005, 07:55
So you have two hot pixels. Not bad. I wouldn't worry about it. As for the question regarding CCDs getting worse over time, there may be a case for that, since I've seen it on some older cameras. Mostly will become a "stuck" pixel, instead of a hot pixel.
Can I just try to clarify the difference between "Hot Pixels" and "Stuck" or "Dead" pixels. I had originally said I saw 4 (actually 3) "Hot Pixels" on my A4 prints. But these are in the same place each time, are red, and just become a little more visable (on prints) at higher iso's. So I think what I am describing is "Stuck" pixels. I can see at higher iso's on screen at 100% other "illuminated" pixels that are usually white or green but do not usually re-occure in the same place. I assume these then are "Hot" pixels.

To be honest I'm not too bothered about these as they are not really obvious in prints and for me that's what counts. I would be concerned if these developed as in more "stuck" pixels but feel reassured by Brian's experience with early Kodak sensors still in use.

JlW's photoshop suggestion as a means of retouching sounds very useful.

Jim