I have owned my trusty Rebel XT for almost five years now. Lately I've noticed that there seem to be more hot pixels than I'm used to when I take photos in low light situations. I decided to perform a test to compare the number of hot pixels that my camera has now, compared to when it was brand new five years ago.
In this test I've included both "hot" and "stuck" pixels. "Hot" pixels are white, and they usually occur in tiny clusters of five on Canon cameras. "Stuck" pixels can be red, green, or blue, and they occur all by themselves, but due to the way colours are processed in the camera, they usually bleed over to the neighbouring pixels as well.
For this test, I found a long exposure photo that I took in June of 2005 on a very dark night, and separated out the hot and stuck pixels from the background of the photo. This was very easy to do, since I had accidentally picked too small an aperture, and thus the photo was very underexposed. Then, I took a photo with the exact same settings and with the lens cap on, so that I could compare the camera today to the way it was almost five years ago.
I then processed both images in the same way, and placed them together. In this weeks photo, you can see the result. On the top is a map of the hot and stuck pixels on my camera when it was only a month old, in June 2005. On the bottom is a map of the hot and stuck pixels on my camera almost five years later, in March 2010. As you can see, there are a LOT more "bad" pixels now... and whenever I take low light photos, this has really increased the amount of work I have to do in Photoshop to clean up the images.
Is it time to buy a new camera? I don't know. Most of the time, for short exposures in the daytime, these hot pixels aren't even noticeable... but for longer exposures in nighttime shots, they are a real pain. In any case, it was an interesting experiment.
Technical details: Both photos were taken with my Rebel XT + kit lens at 18mm, ISO 100, f/18 for 235 seconds. Both were processed by adjusting the levels to remove the background, and then running a threshold filter set to 5 to clearly identify the bad pixels, and then scaled down for the web.
That isn't uncommon (for hot/stuck pixels to appear over time). Are you using Adobe Camera Raw for the conversion? I ask because ACR does a bunch of hot pixel removal automatically. You should try the same experiment in DPP and see if the results are any different (make sure you turn down the color and luminance noise reduction though). If you want to be really pedantic, you should download dcraw (and source code) to get the least molested debayered image. You can then do all kinds of fun geekery like write up your own image statistics to quantify the number of hot/stuck pixels. Fun stuff...-- Aravind at 12:45pm, Friday March 12, 2010 EST
Cool! I didn't know there was software out there that would give you the debayered image. Yeah, in all honesty, I just took the picture, converted it to TIF in Raw Image Task, and did some quick fanagling in Photoshop. I didn't spend too much time pixel peeping. :)-- Michael at 8:37pm, Friday March 12, 2010 EST
Check this out: http://www.naturescapes.net/phpBB3/viewtopic.php?f=2&t=173769 Try that and see if the number of hot/stuck pixels goes down.-- Aravind at 12:27pm, Tuesday March 16, 2010 EST
Actually I tried that already... I didn't notice any difference at all, even though other people swear it works on the Rebel XT. :(-- Michael at 3:10pm, Tuesday March 16, 2010 EST
Dang, thats too bad. I'm going to try it with my 1Ds3 and see what it does, thats the body I have with the most hot pixels (though none make it through ACR's filtering).-- Aravind at 8:38pm, Tuesday March 16, 2010 EST
Hosted by theorem.ca