3
\$\begingroup\$

Recently I noticed the appearance of a quite big spot of increased brightness and strange colour in my longer exposure photos. I'm using a Samsung NX300m (I know, not a good camera for low light...).

As an example, this was shot last night (15 sec. exposure, ISO 3200). I made post changes in post — no brightness adjustments, no LR Noise Reduction, camera's "long exposure noise reduction" was disabled, too.

real shot

Notice the slightly brighter and purple-ish spot in the lower left corner. To prove that this is not a problem of the lens, I just took another shot, this time with the lens cap on and in a dark room. 25 sec. exposure, ISO 200, no noise reduction, heavily boosted in Lightroom.

noise shot

Is anybody able to tell me what is happening here? Did I damage my sensor? Was the sensor just really, really bad for everything longer than 10 seconds all along? And most importantly, can I do anything about it, hopefully other than buying a new camera?

Thanks a lot!


Edit: (seems I'm not allowed to comment directly to your answers since I'm new to the community.. Already, thank you for your answers!)

I recreated the second Photo from above (lens cap on, 25s) only this time with the internal LENR feature turned on as suggested.

LENR on this time

As expected from what the answeres say, the big spot and the nasty green line, a.k.a. fixed-pattern-noise got a little better. But therefor, the overall (random) noise got a lot worse! (LR brightness boost is the same in both images).

My own guess: The original random noise and the noise the sensor generated while taking the dark frame kind of added up, since, well, it is in fact random, and taking two samples is not enough to obtain a good average value.

Now, when shooting fixed, dark scenes, would I be better off creating my own dark frame from multiple exposures and substracting that from the original image?

\$\endgroup\$
1

2 Answers 2

1
\$\begingroup\$

First, let's distinguish two kinds of noise. One is temporal random noise, such as thermal noise, photon shot noise, or electron shot noise in the analog electronics. These tend to follow Poisson distributions, and vary from shot to shot. Since Poisson -based signal-noise ratio goes as the square root of total signal, the noise tends to be observable only at small signal levels.
The other kind of noise is spatial non-uniformity of response, often called "fixed-pattern noise." This can be due to a variation in quantum efficiency across the sensor region, variation in analog gain, etc., and manifests itself as a non-uniform mean exposure level.
Your image strongly suggests the latter, i.e. the bright region is fixed-pattern noise. If properly implemented, a "long-exposure noise reduction" algorithm corrects this, as Linwood said, by taking a known dark image and subtracting it from the exposed image.

\$\endgroup\$
1
  • \$\begingroup\$ I did another noise shot, same settings, only this time with LENR turned on. i.sstatic.net/ngPSf.jpg Fixed-pattern-noise improved. But the random noise is a lot worse! I'm guessing, the noise from the two exposures (light/dark) added up, since it is of course random, and taking two samples is not enough to obtain a good average value. Now, when shooting fixed, dark scenes, would I be better off creating my own dark frame from multiple exposures and substracting that from the original image? \$\endgroup\$
    – smow
    Commented Nov 21, 2016 at 10:39
5
\$\begingroup\$

I do not know that specific camera, but generally sensors will have a pattern of thermal noise which is not uniform. The "Long Exposure Noise Reduction" feature, if your camera has it, is intended to offset this somewhat. Enabled, it does much as you did, taking one image with the shutter open, and one with the shutter (or in your case the lens cap) shut. In camera it would then subtract, numerically, each pixel of the latter from the former. The process is not perfect; noise is statistical in nature, and not perfectly repeatable. By that I mean that a specific pixel will not have the same value from such a "dark" frame twice, but what does happen is that some areas tends to be brighter, repeatably, in the same spot when taken under the same condition. Doing this "dark frame" subtraction thus selectively darkens the area.

One way to think of it, which is not far from reality, is that each pixel is a well that becomes more full as it exposes. Noise, in a sense, is garbage that is in the well and artificially fills it somewhat. The dark frame adjusts each pixel trying to get it back to an empty starting point. It is much better than just darkening the area, as that is proportional -- you are darkening the garbage instead of removing it, in a sense.

The dark frame subtraction (aka Long Exposure Noise Reduction) is also one of the few settings that affects a "raw" image in most cameras, high ISO noise exposure, for example, is not applied to the raw image. Long exposure noise reduction is.

Try turning on Long Exposure Noise Reduction. If you are curious about the technique, including being able to apply it manually, search for astrophotography and "dark frames", it is a very common practice there. It is also useful to do manually if your camera does not support it in camera, or if you need a lot of images in the same conditions and do not want to wait for a 2nd exposure for each one.

To your original question - did something you did cause this to be worse? My guess is no, but without some before and after shots, no way to know for sure. But what you posted does not look that unusual. Some cameras are better than others at producing less noise, but all are going to have some sort of pattern like this if you push hard enough to see it.

\$\endgroup\$

Your Answer

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

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