10

After shooting for one hour with my Rebelt2i + Tamron 17-55 f2.8, I went back home and checked pictures on my computer.

For every single picture, there are a lot of 'dead' pixels at the same location. I searched on Google and it could be dust, dead pixels, hot pixels... Do you have any suggestions on what the problem really is, and how to solve it?

Look between two buildings on the right, a few in the sky too: sample image (Click for large version.)

Question also asked by gsharp:

Every picture that I took today have a red dot on the same place in every photo. It looks like a monitor "pixel error". Is just the lens dirty or did something bad happen to my cam?

Here some samples. Best mode to detect the dot is download the picture in original size (click the (i) then download)

Community
  • 1
Eno
  • 267
  • 1
  • 3
  • 4
  • Just to confirm; this happens on regular basis not only this one time and with this particular image? – Jakub Sisak GeoGraphics Aug 14 '11 at 11:50
  • It happends only for long exposure ( +20sec ), that means every pictures from that night have hot pixels at the same location. – Eno Aug 14 '11 at 18:16
  • 9
    Possible Explanation: There Really was a Red Dot in Every Scene

    The question is: are you following the Red Dot around, or is he (they are usually male) following you around?

    Red Dots love to be photographed and are known for jumping out on photographers and posing in random shots.

    I'll grant that this is an unlikely explanation.

     – AJ Finch Oct 26 '11 at 11:22
  • 7
    @AJFinch - Obligatory Mitch Hedberg quote: "I think Bigfoot is blurry, that's the problem. It's not the photographer's fault. Bigfoot is blurry, and that's extra scary to me. There's a large, out-of-focus monster roaming the countryside." – anon Oct 26 '11 at 20:59
  • 2
    @anon. Perhaps I should try tracking down Bigfoot; I'm very skilled at photographing blurry people, as my photo collection demonstrates – AJ Finch Oct 28 '11 at 11:58
  • lol @ AJFinch :-) – gsharp Oct 28 '11 at 13:06

5 Answers5

15

You're exactly right — it's the same thing as a monitor pixel error, but on your camera's sensor rather than on an LCD screen. *

You can either fix it in post-processing (automatically, with many RAW-processing packages) or have it mapped out with the camera's firmware.

If you're lucky, your camera model includes a built-in feature to do that yourself in the field; failing that, it's almost always covered under warranty service.

* well, not really exactly the same, since it's a photoreceptor rather than an LCD pixel, but it's similar in concept. One of the image elements is malfunctioing.

mattdm
  • 143,140
  • 52
  • 417
  • 741
  • oh lord. is this something normal/usual or do I have a defect cam? – gsharp Oct 25 '11 at 22:12
  • It is fairly common. And consider, if you have one dead pixel on a 10megapixel sensor, that's a 0.0001% loss. No big deal, as long as you mask it so it isn't annoyingly bright red. – mattdm Oct 25 '11 at 22:16
  • I guess it affects more that one pixel, because the dot is visible in every photo in original size. – gsharp Oct 26 '11 at 09:10
  • 1
    @gsharp — Can you post an example? A single stuck-on red photosite can really stick out, particularly since it will also affect the surrounding pixels in the Bayer filter interpolation (which reconstructs a full-color image from a bunch of red, green, and blue photosites). If you have the photosite masked out, that won't happen; it'll just be ignored and you won't notice it at all. But it is possible that you have a cluster of dead pixels, or some other problem. Posting an example will certainly help us figure out what's going on. – mattdm Oct 26 '11 at 19:36
  • posted some samples --> see question. thanks for your time and effort. – gsharp Oct 28 '11 at 15:41
  • 2
    @gsharp try shooting a 2 minute long exposure with the lens cap on and watch the resulting christmas tree. It is really common to have cold/hot pixels on a camera sensor. It shouldn't be so apparent on sub-second exposures, but it is not something to worry about. Clone it away, tell the firmware to ignore it, whatever. It's just a single pixel. – pau.estalella Dec 27 '11 at 16:37
5

This effect is a very common one and is caused by dead or damaged pixels.
Pixels may be "stuck on" and a single pixel producing a bright one colour dot or a small group of pixels

In images they are often termed "hot pixels. Most sensors have some. They are a common consequence of aging. Some sensors will have them from new, some won't.

LCD screens also have them. Software is available to map them and to adjust the image to greatly remove their effect. As a simplistic example, if you had a single "hot pixel" if you averaged the light values in the adjacent pixels and replaced the "stuck" value wit the average value it would be exceeding unlikely to be noticed in the very large majority of images. The algorithms used for correcting them are liable to be more complex than this but this explains the concept well enough.

Some cameras have a hot pixel correction function - a long exposure shot is made in 'total darkness' (eg - with lens cap on) and the camera then adjusts for anything it finds in the dark frame.

Hot pixels will usually be most obvious in areas where they stand out against the background, and in longer exposure images.

"Stuck" pixels are most obvious if they are stuck on - but pixels that are always off will also be encountered. Here's an image showing an always off green pixel.

enter image description here

Here's a comment on them by Ken Rockwell

Wikipedia - defective pixels

Also informative

DPReview user discussion

Think yourself lucky

Related

Russell McMahon
  • 19,630
  • 1
  • 44
  • 79
  • What is this caused by? How does the damage occur? Also, why are there several bright pixels in one dot instead of a single pixel? Are there actually several adjacent damaged pixels on the sensor, or is this effect due to demosaicing? – Szabolcs Aug 08 '12 at 11:24
  • Read the references. Look at the large amounts that Gargoyle says on the topic and then ask further if desired. – Russell McMahon Aug 08 '12 at 11:36
  • Hey Russell, could you move this answer to http://photo.stackexchange.com/questions/14889/why-is-my-camera-showing-a-lot-of-stuck-red-and-white-pixels? – mattdm Aug 08 '12 at 11:56
  • @mattdm - I can just reenter the answer details under the other question but that doesn't "move" it per se. Is there a formal means of moving it? Points from upvotes are not especially important wrt their affect on my reputation BUT the upvote count against the answer is an indication of the value that others thought the answer has, and I do value that. – Russell McMahon Aug 08 '12 at 18:13
  • A mod can move it, but after it's accepted as an answer the only way is by doing merge of the questions. I'm not saying it's a bad answer — it is — but it's really better for people looking for information later if we can consolidate all basically-identical questions in a canonical place, with all the answers there. And this one we get a lot! – mattdm Aug 08 '12 at 19:57
  • @mattdm - Did your "it is" mean "it isn't" or did "it is" mean "it is" ? :-) – Russell McMahon Aug 08 '12 at 22:48
  • :) Let me try that again. "I'm not saying it is a bad answer; to the contrary, it's a good answer. But...." – mattdm Aug 08 '12 at 22:54
1

Those are probably hot pixels, to test this just take a picture with the lens cap on - this picture will come out all black except for hot pixels (dust or lens problems wouldn't show up because there's no light to see them with).

I have no idea what's the acceptable number of hot pixels and when you should send the camera to be fixed but I can tell you how to map out those pixels, this worked for my 550D (copied from my answer on another question)

http://www.slashgear.com/how-to-remove-stuck-pixel-on-your-digital-slr-2227392/

The relevant part:

Test for dead pixels : Left lens-cap on, set camera to exposure 30 seconds black-out image at various ISO settings.

Fix for dead Pixels : Set camera to perform sensor cleaning mode manually (not auto) for one minute. You only execute the function, not exactly performing the whole sensor cleaning procedure. Magically, the sensor cleaning task should shake the stuck pixels off, if not, repeat couple more times until it does. Otherwise, time to call the technical support.

I've tested for hot pixels using only ISO 100 (I'm lazy), the image clearly showed the 2 pixels that caused me to look for this + another red pixel.

After enabling the manual sensor cleaning function (without even taking the lens off) and waiting for a minute the first two pixels disappeared, after 2 more "cleanings" the third also disappeared.

Imre
  • 31,966
  • 11
  • 107
  • 177
Nir
  • 20,825
  • 4
  • 38
  • 74
  • Note that the sensor cleaning task does not actually "shake the stuck pixels off", as the linked text suggest. It apparently is doing the separate operation of mapping out the hot pixels. On many cameras (Olympus and Pentax, at least), this is a separate menu option. It is a little bit odd, though, that the suggestion is to repeat several times. That should not be necessary. – mattdm Aug 14 '11 at 13:41
  • @mattdm - I agree that in theory repeating several times should not be necessary - but in reality, when I did it with with my actual camera, it took 3 times until all the hot pixels were mapped out. – Nir Aug 14 '11 at 14:56
  • 3
    The sensor cleaning mode is purely a myth. As far as I know it is not linked to any pixel mapping on any Canon cameras. Any higher end cameras that have have a hot pixel mapping option have it separately in a menu option, and any lower end cameras that don't I think you can send them off to have it done. There may be a way to do it with software that I don't know of. All the sensor cleaning option will do after seeing 'Hot pixels' is open up the sensor to the air, which will cool it down slightly quicker, meaning the next test shot may have less pixels visable. – Dreamager Aug 14 '11 at 17:42
  • ^^^^ This is nothing more than confirmation bias – Dreamager Aug 14 '11 at 17:42
  • And to add onto that, I've the same problem on my 500D. There are a few pixels that are very noisy at 1/8s or slower, and some others that show up when it's more than a few seconds (on ISO100). Anything quicker is fine. It was finding fixes to this that I came across the confirmation bias myth. I 'think' there's a way in photoshop and other programs to map out the hot pixels from a test shot, but that seemed complicated, so instead on PSPX3 I just use the spot remover tool to get rid of them if it's a picture I want perfect. – Dreamager Aug 14 '11 at 17:55
  • I tried the 'quick' fix with the manual cleaning and it didn't work, how many 'dead' pixels are acceptable for a camera? This is a full black 30sec exposure picture. http://dl.static-storm.com/black.JPG I can see at least 10 dead / hot pixels. – Eno Aug 14 '11 at 18:15
  • That doesn't seem too overkill. I've a cluster of about 4 together in one spot that's quite annoying. I think that amount is about par for the course with long exposures though. Nothing to worry about, but I'm not sure on the best fix – Dreamager Aug 14 '11 at 19:38
  • Just to completely debunk the sensor cleaning myth, if it did map anything while in that mode, whilst actually cleaning the sensor the whole sensor would be detected as a dead pixel as everything would be whited out from the amount of light hitting it. There is zero power to the sensor whilst in sensor cleaning mode, as otherwise it would be electrostatic, and attract more dust than was previously on there before cleaning, completely defeating the purpose of the task. The sensor must remain powerless to be cleaned. – Dreamager Aug 14 '11 at 19:41
  • @Dreamager - when I've found the hot pixels I've started looking at older photos, those pixels have been there for weeks - and I did the cleaning thing without taking the lens off - so your cooling off theory is incorrect (if anything I've heated it up). now, I don't think the sensor is active during cleaning but the camera could map the bad pixels when coming out of cleaning mode after closing the shutter -- sorry but you are not debunking anything, you are just throwing around theories like the rest of us – Nir Aug 14 '11 at 20:13
  • @Dreamager - if you know a Canon engineer with access to the firmware source code or you know how to take care of hot pixels in Canon entry level models I'd love to hear it - I don't claim I know everything - but saying two things are not connected is not debunking - after all it's just software and it could do whatever it wants before or after cleaning – Nir Aug 14 '11 at 20:15
  • Sadly I've no direct access to any Canon resources or techs, and would have to go scowling the net again to link to the proof of the myth (I may do later). With regards to the cooling off, you'd not need the lens off. If you do it just after you've taken some test shots, leaving the sensor off and open to air of the camera for 30s will cool it quicker than having the shutter closed. – Dreamager Aug 14 '11 at 21:21
  • And with regards to pixels you've noticed for ages, yes it's likely they have been there for ages, and on those pictures you're seeing them you'll have been using your camera for a little while and internals will have heated up. Doing the 'sensor cleaning' and then looking again for the pixels will have made sure the sensor is cool so there may well be less showing. My most noticeable pixels appear after less than a second, so the 'sensor cleaning' has no effect on them. – Dreamager Aug 14 '11 at 21:24
  • 1
    If after the 'sensor cleaning' the software was to map the pixels as you suggest, as the cleaning left the sensor off for 30s (or however long you left it open), would the camera not then have to spend 30s for the sensor to expose itself so that it could see the pixels it needs to map? – Dreamager Aug 14 '11 at 21:26
  • @Dreamager - Yes, and (at least on canons) it does exactly that. The behaviour is controlled by a custom-function: "Long exp. noise reduction". With the setting on, any exposure over a threshold for long exposure (I think it's a few seconds) takes twice the time the shutter is open to complete. A 30 sec exposure takes 1 minute. – Fake Name Aug 15 '11 at 05:40
  • 1
    @Fake Name Yes I gave that a mention in my answer, it's very effective I hear although 'cos of the double time I've not used it myself. This extra time is something that definitely doesn't happen after the 'sensor cleaning' myth method however, so when the pixel mapping is supposedly supposed to take place I don't know. – Dreamager Aug 15 '11 at 09:25
1

After moaning about the supposed pixel mapping myth in the comments I thought I'd provide one possible solution. Digital Photo Professional provides a dust mapping option for dirty sensors that removes dust data from images. Info about it can be found here http://support-th.canon-asia.com/contents/TH/EN/8201038200.html

I 'think' you can use this method with test images of the hot pixels to automatically remove those locations from the image, but I haven't tried it myself and wouldn't exactly be sure how to do this, so this is only really a half answer. Maybe one worth exploring though.

Edit: Also, depending on how long your exposures were at that time, if you use Long Exposure Noise Reduction, the camera will take a hot pixel map after your photo for the same length as the exposure to remove hot pixels. This may not be convenient though with the extra time it takes, especially with very long exposures. Also I'm not sure the minimum exposure lenth it needs to be to kick in.

Dreamager
  • 2,215
  • 1
  • 19
  • 18
  • 1
    Adobe Lightroom does (effective) automatic hot-pixel removal. DXO Opticsworks does it too, though it's significantly less effective. Neither needs a dark-frame. – Fake Name Aug 15 '11 at 05:36
  • The dust data method is another mathematical model than is needed for hot/dead pixels, so that part of the answer is wrong, but the long exposure noise method is a nice bit of information. – Michael Nielsen Dec 20 '12 at 10:04
0

When shooting raw, hot pixels are often handled by telling the raw converter to ignore signal from pixels listed in dead pixels file or similar. When shooting JPEG it's get harder, because it's camera, not PC software that does the interpolation step.

In normal digital cameras the raw conversion needs to interpolate pixels even if every sensor pixel was working, because you get only one color component per pixel.

Zds
  • 1,489
  • 9
  • 11
  • The camera can be made to ignore the stuck pixels in its JPEG conversion. In most high-end cameras (and low-end ones from at least Olympus and Pentax; maybe others) you can do this from a "map out hot pixels" menu item. For others, it can be done at the factory, almost always as free warranty service. – mattdm Aug 15 '11 at 12:54
  • Yeah, that's why I said "harder", because you need manufacturer support to get it done, so the easiness varies. – Zds Aug 15 '11 at 13:22