0
\$\begingroup\$

I took photos at two separate fight events in different venues with similar settings and a huge difference in noise. The ring lights in a dark venue aren't always the best especially for high shutter speeds to freeze the action.

Sony a7iii first fight event shot with settings at
28-70mm, f/3.5-f/4.5, 1/800, 10000 ISO almost no noise. Second fight event settings 28-70mm, f/3.5-f/4.5, 1/640, 8000 ISO extremely noisy. Speed burst and no flash.

The only difference is lighting and venue and second event I had a lens hood on. I want to avoid future noise in photos obviously but can't figure out why this happened. The image below is a side-by-side comparison.

enter image description here

Original image link

\$\endgroup\$
3
  • \$\begingroup\$ You may find this question helpful: What is the physical cause of increasing noise at high ISO? and especially this answer photo.stackexchange.com/a/131840 \$\endgroup\$
    – wonderbear
    Jun 4 at 15:29
  • \$\begingroup\$ As others pointed out, one would have to see the unprocessed images, or at least the complete metadata of the images. I also suspect that the temperature of the sensor and electronics could contribute to the noise: Was the camera on all the time, and which one was taken first (assuming it was from one session)? \$\endgroup\$
    – U. Windl
    Jun 4 at 23:14
  • 1
    \$\begingroup\$ @U.Windl It's the light! It's almost always the light. It certainly is here. 1 - The subjects are being front-lit in front of a dark background. Virtually all of the bright light is on the subjects. 2 - There's a spotlight above the left glove of the boxer in red that is the majority of the light in the scene. The subjects are backlit (look at their shadows) and barely brighter than the background. The camera reduced exposure due to the bright spot, otherwise it would have metered the scene 3-4 stops darker. \$\endgroup\$
    – Michael C
    Jun 8 at 2:50

2 Answers 2

7
\$\begingroup\$

Only difference is lighting and venue...

That's all the difference you need.

Addendum: Now that you've posted example images, we can point out some other differences as well.

In the earlier image on the left, there's far more light falling on your subjects than in the image on the right. At the same time, there's far less light in the background. The brightest things in the frame are the two contestants. All of the noise in the background is swallowed up by the dark background being reduced to pure black by your camera's processing engine (or your raw convertor's processing engine).

In the image on the right, there's far less light falling on your subjects, at least from the direction the camera is pointing. But there's more light in the background. The spotlight in the middle of the frame above the left glove of the boxer in red is causing the camera to meter the scene much brighter than it otherwise would. Instead of exposing for the subjects, as it did in the first example, it's exposing to not blow out the spotlight in the second. Your subjects are much dimmer in the second image than the first. Other than the spotlight, the background light is still pretty weak, but it's too strong to show up as pure black against your dimly lit subjects that aren't much brighter than the background. So all of the noise in those dimmer areas gets amplified and shows up in the darker, but not totally dark, parts of the image.

Notice the shadows of the boxers? Most of the light on them is coming from the other side of the ring. In other words, the sides of the boxers away from the camera are brighter than the sides of the boxers in view of the camera. If you'd shot from the other side of the ring, at least your subjects would have been brighter and less noisy compared to the background.

Noise is not a direct function of what ISO setting you used. Many people believe that it is, but it isn't.

Noise is a function of the ratio of usable signal, that is light, to the amount of random signal and signal not created by light entering the camera and being recorded by the camera's sensor.

In other words, the ultimate determiner of how noisy an image looks is the signal-to-noise ratio, a/k/a SNR or S/N ratio. Signal when talking about digital camera sensors is just another word for light. The more light you allow into the camera, the more signal you have. The less light you allow into the camera, the less signal you have.

How much light you allow into the camera is determined by three things:

  • The entrance pupil of the lens a/k/a the "aperture"
  • The length of the exposure a/k/a the "shutter speed"
  • The intensity of light falling onto the front of the lens from within the camera's field of view a/k/a "scene brightness"

The more light there is landing on the front of the lens, the more signal will enter the camera at the same aperture (Av) and exposure time (Tv). When the scene is dimmer we must either open up the aperture or lengthen the exposure time to allow the same amount of light into the camera.

Notice that ISO doesn't appear anywhere in the determination of how much signal the sensor collects. ISO affects how much electrical amplification is applied to the signal after the sensor collects it. But this electrical amplification also applies to any noise already included in the sensor's output at the time the signal is amplified.

Noise has two basic sources:

  • The random nature of the distribution of photons within light itself
  • Electrical current in the camera's electronics that is falsely measured as signal

The latter is generally constant over a given time interval for the same sensor operating at the same temperature. If the sensor gets warmer, it also tends to be noisier due to this "dark current".

The former is due to the way photons aren't uniformly distributed in light from light sources. If the intensity is very weak, the randomness is more noticeable in short exposure times, and the image will be noisier. If intensity of light is very weak, a longer exposure that collects more light will average out some of the randomness of the weak signal. If the intensity is strong, the larger number of photons streaming from the light source will average the randomness out more quickly.

The amount of noise due to this randomness within light is known as Poisson distribution noise, after the name of the mathematician who first introduced the concept of discrete probability distribution that may be observed in a wide variety of applications. Poisson distribution noise increases at the square root of the intensity of the light field. We also often refer to Poisson distribution noise as "shot" noise. So as the intensity of light increases, so does the shot noise! But since the intensity increases as the square of the randomness, the overall result is that the shot noise gets more and more drowned out by the sheer amount of light as the amount of light increases.

So what makes an image noisy?

Not enough light collected over too short an exposure through too narrow an opening.

If you shot both events at the same apertures and more or less the same exposure times, then the difference was that the venue for one event had more light on your subjects than the other. Most of the light in the second image is concentrated at the spotlight in the field of view. Everything else is several stops darker.

\$\endgroup\$
9
  • \$\begingroup\$ The light levels on the subjects are essentially the same in both images... the first image used a slightly smaller aperture, slightly faster SS, and slightly higher ISO. I.e. the SNR of the scene appears to be about equivalent, and less light actually recorded in the lower noise example. \$\endgroup\$ Jun 4 at 21:16
  • 1
    \$\begingroup\$ @StevenKersting The same meter reading does not always mean there's the same amount of total light in the scene. Matrix/Evaluative/whatever Sony calls it metering means highlights have more influence over metering than total light. Even when the same total amount of light is recorded, having one image with areas that are all either very bright or pure black will be less noisy than another that has most areas somewhere in between the two extremes. \$\endgroup\$
    – Michael C
    Jun 6 at 0:24
  • \$\begingroup\$ @StevenKersting In other words, even if total brightness of the FoV is the same, the light level on the subjects is not "the same" for both images. One is high key, one is low key. If the second had been taken from the other side of the ring, it might have looked more like the first. We can see by the shadows being cast by those in the ring that they are being lit brighter from the other side of the ring than from the camera side. \$\endgroup\$
    – Michael C
    Jun 6 at 0:26
  • \$\begingroup\$ @StevenKersting Not to mention that the spotlight inside the FoV in the second image greatly reduced the metered exposure. Without that light in the frame, the camera would have metered much dimmer and thus exposed much brighter (if allowed to). Most of that "equal light" in the second scene is concentrated in a small circle above the left glove of the boxer in red. \$\endgroup\$
    – Michael C
    Jun 6 at 0:30
  • \$\begingroup\$ if the exposure of the subjects is equally bright (and it is) and the exposure settings were about equivalent (they were), then the amount of light on/from the subjects is roughly the same. If the light in the second image BG had significantly affected the exposure, then it would not have blown out completely as it did. The noise level will vary throughout an image as the SNR varies throughout... the only way these results make any sense is if the images are heavily/differently edited. \$\endgroup\$ Jun 20 at 16:31
0
\$\begingroup\$

I'm struggling to accept these images as not being edited with significant exposure/shadow recovery or other contributing factor involved.

ISO in itself does not cause noise; if anything it reduces noise. If those images are unedited then the noise levels would be at least as high in the "low noise" example because even less light was recorded. Much of the noise would not be visible do the much greater contrast (hidden in shadow/blacks), but I don't think that fully accounts for the differences shown.

\$\endgroup\$
3
  • 1
    \$\begingroup\$ It's pretty obvious most of the light in the second image is not on the subjects, it's the spotlight above the left glove of the boxer in red that is backlighting the boxers. The subjects themselves, as seen from the camera, are lit significantly brighter in the first image, where they are by far the brightest thing in the scene. In the second image, they are barely brighter than the dim background. The spotlight shining straight at the camera influenced the meter and caused the camera to expose darker than what it would have had the spotlight not been in the FoV. \$\endgroup\$
    – Michael C
    Jun 6 at 0:45
  • 1
    \$\begingroup\$ In the first image, practically ALL of the light is on the subjects. In the second image Most of the light is not on the subjects, but coming from the spotlight. The subjects are much darker in the second image than in the first because the camera was exposing for the subjects in the first and exposing for the spotlight in the second. Without the spotlight the camera would have metered at least 3 stops, maybe more, differently. But it wouldn't have mattered. The subjects aren't significantly brighter than anything else from that angle. See the light! \$\endgroup\$
    – Michael C
    Jun 6 at 0:50
  • 2
    \$\begingroup\$ If you look carefully at the File Path shown, the files are Sony ARW RAW files. So yes they are edited and you can see significant exposure/shadow recovery in the black around the fighters in the first image. \$\endgroup\$ Jun 9 at 18:49

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.