9
\$\begingroup\$

What is the difference between hot, stuck, and dead pixels with regards to a camera's image sensor? What can cause each to occur? What can be done to reduce their influence on a photo?

\$\endgroup\$

3 Answers 3

16
\$\begingroup\$

Excitingly, these terms mean different things to different people. I think the most useful distinction is like this:

  • Stuck pixels are always completely bright, as if they're fully overexposed
  • Dead pixels are always off, as if receiving no light (these are usually less obvious)
  • Hot pixels are not permanently stuck, but show up during long exposures (as the sensor heats up). These usually are defective pixels to some degree, and the same sensor will usually have the same hot pixels in the same conditions.

Some people use "hot" and "stuck" interchangeably, for either one or both of the different situations. I think it's more useful to make a distinction, but be aware that not everyone does — it's most helpful to be a little more verbose and explain what exactly you mean.

"Dark frame subtraction" is an effective way to deal with hot pixels. The same can be done for stuck (or dead) pixels, but generally the thing to do is have them mapped out permanently. Many cameras have a function in the menu to scan for these defects and do the mapping; otherwise, it's usually covered under warranty service. (This may or may not also get some or all of the hot pixels.)

In a digital sensor, all of these usually refer to a single photosite, with its corresponding color filter, so you might have a bright spot of red, green, or blue. This may also "bleed" into nearby pixels during the demosasicing process (but only as an artifact, not from electrical leakage or anything), resulting in + or × patterns.

\$\endgroup\$
10
  • 1
    \$\begingroup\$ +1 for "excitingly" and the unrestrained glee it instilled in me while reading this. \$\endgroup\$
    – Jason C
    Feb 25, 2017 at 0:50
  • 2
    \$\begingroup\$ @JasonC Okay, I admit, my perspective on what exactly constitutes "excitement" may be skewed. \$\endgroup\$
    – mattdm
    Feb 25, 2017 at 3:13
  • 3
    \$\begingroup\$ @Ruslan In principle, yes. But a) dead pixels tend to be less obvious. an d b), I suppose that at least some cameras replace hot/stuck pixels recognized by dark frame subtracton not by black but rather by the average of their neighbours or the like. \$\endgroup\$ Feb 25, 2017 at 9:19
  • 1
    \$\begingroup\$ @HagenvonEitzen For "hot" pixels that aren't saturated, the subtraction is, at least theoretically, a proper correction. Another reason to distinguish between "hot" and "stuck". \$\endgroup\$ Feb 26, 2017 at 0:22
  • 1
    \$\begingroup\$ Well, yes, it's all "noise" as in "not signal", but... hot-sensor bright pixels in a dark frame, which are notable outliers from the normal distribution of thermal entropy, even if not at full voltage, would not be "hot pixels" unless they hit saturation? Just trying to clarify the use of terms... \$\endgroup\$ Feb 26, 2017 at 21:15
3
\$\begingroup\$

Hot pixels are pixels that read out at full voltage/saturation as a result of heat in the sensor. Once the sensor is allowed to cool they will usually return to their proper state and read out a voltage value based on the amount of light that has struck the pixel well the next time they are used. They most often present an issue during long and/or high ISO exposures. Ambient temperature can also influence the number and severity of hot pixels in a photo. The same conditions that typically produce more digital noise also typically produce more hot pixels.

Often the best way to deal with hot pixels in a captured file is to use dark frame subtraction. One can also manually use a dust repair tool or healing brush tool to edit out hot pixels. In order to prevent occurrence in future captures one can allow the camera and its sensor to cool.

Stuck pixels are similar to hot pixels in that they read out at full voltage/saturation but they do it all of the time. That is, they are permanently stuck in the full "on" position.

Dark frame subtraction can also deal with stuck pixels, but a more effective and convenient way is to use pixel mapping. Some cameras do pixel mapping automatically. But if pixel mapping is done when the sensor is warm or hot, then hot pixels can be mistaken for stuck pixels and be mapped out of subsequent shots even after the sensor has cooled. In Canon cameras, for instance, the sensor is remapped each time the firmware is updated and each time the shutter is held open for at least 30 seconds using the manual cleaning function. Recording dust delete data is an effective way to map out stuck pixels with some cameras.

Hot or stuck pixels can often influence the values of the surrounding pixels as well. When the camera or RAW converter program converts the information from the sensor, the high value of that one pixel can also cause the surrounding pixels to have artificially high color values (for the color of the hot pixel) assigned by the interpolation of the demosaicing algorithms.

Dead pixels are the opposite of stuck pixels. They are permanently stuck in the full "off" position. Cameras will often interpolate values for dead pixels from the surrounding pixels in much the same way they interpolate color values for all pixels from the monochromatic values in the raw data.

What can be done over the long term to reduce the probability of hot, stuck, or dead pixels?

Only energize and expose the sensor to light when necessary. The more light and electrical energy a sensor is exposed to for a longer period of time the greater the number of pixels that develop defects. The number of stuck, hot, or dead pixels increase over the life of any sensor. Leaving the mechanical shutter open and the sensor energized and exposed to light over the course of an entire shooting session will reduce the life expectancy of the sensor by a considerable amount compared to using a mechanical shutter and only energizing and exposing the sensor to light when actually recording an image.

\$\endgroup\$
12
  • 2
    \$\begingroup\$ Hmmm. I'd like to see a citation on the part about increased degradation with exposue (as opposed to simply over time, or with time at higher temperatures). \$\endgroup\$
    – mattdm
    Feb 25, 2017 at 2:02
  • \$\begingroup\$ Ask any electrical engineer what happens to a conductor or semiconductor when current flows through it. It's pretty basic physics. The smaller integrated circuitry becomes, the more it becomes a concern. At the other end of the extreme the extremely high current flowing through them is why high tension wires have to be replaced periodically. en.wikipedia.org/wiki/Electromigration \$\endgroup\$
    – Michael C
    Feb 25, 2017 at 2:11
  • \$\begingroup\$ spectrum.ieee.org/semiconductors/processors/transistor-aging \$\endgroup\$
    – Michael C
    Feb 25, 2017 at 2:25
  • 5
    \$\begingroup\$ Yeah, I don't doubt that that's a thing. I'm just wondering if it's really a significant factor over the lifetime of a DSLR or mirrorless camera, particularly in the context of other factors. \$\endgroup\$
    – mattdm
    Feb 25, 2017 at 3:15
  • \$\begingroup\$ There are other issues as well that microcircuits experience due to being energized. And then there are gamma rays (which aren't dependent upon an IC being energized). Mission critical applications that use integrated circuits are multi redundant to combat this and still have rated service lives that are shorter than their non-mission critical cousins that do similar tasks. \$\endgroup\$
    – Michael C
    Feb 25, 2017 at 3:18
1
\$\begingroup\$

What is the difference between hot, stuck, and dead pixels?

Dead pixels are stuck in the off state (totally dark), while hot pixels are stuck in the on state (they stay on).

What can cause each to occur?

Hardware problems. You could think of a dead pixel on a video screen like a blown light bulb -- it just doesn't work any more. Image sensors are sort of the opposite of video screens in that they absorb light and report the brightness instead of converting some brightness value to light, but it's the same idea: some photosites will just always report a brightness value of 0 (dead), others will always report a different value like 255 (hot).

What can be done to reduce their influence on a photo?

Many cameras can map problem pixels and compensate for them. Check the user manual that came with your camera for the particular procedure, but it's basically like dust mapping: you take a photo of an evenly lit surface like a wall and let the camera find the pixels that stand out. Once it knows where the stuck pixels are, the camera can ignore them and use values from neighboring pixels.

\$\endgroup\$
1
  • 2
    \$\begingroup\$ I think the explanation of hot pixels is inaccurate, as it doesn't explain why they're more of a problem with long exposures. Also, long exposure noise removal via dark frame. \$\endgroup\$
    – ths
    Feb 24, 2017 at 22:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.