Shot noise, also called photon noise, is the random variation in the number of photons that reach each pixel of an image sensor during an exposure. In photography and image quality, it is the dominant source of noise in most well-exposed images, and it comes from the nature of light itself rather than from any flaw in the camera’s electronics.
Because photons arrive at random, two pixels receiving the same average light will never record exactly the same count. The resulting grain is fundamental and unavoidable; it can be reduced only by collecting more light.
How shot noise arises from photon statistics
Shot noise follows Poisson statistics. When a pixel collects on average N photoelectrons, the actual count varies from exposure to exposure with a standard deviation equal to the square root of N. A pixel expecting 10,000 electrons will typically record a value within about 100 electrons of that average, while a pixel expecting 100 electrons varies by about 10.
The key consequence is that signal grows faster than noise. The signal rises in proportion to N, but shot noise rises only with its square root, so the signal-to-noise ratio equals the square root of N when shot noise dominates. More light therefore always means a cleaner image, even though the absolute noise value increases.
| Collected electrons | Shot noise | Signal-to-noise ratio | Visual result |
|---|---|---|---|
| 10 | About 3.2 | About 3.2 | Very grainy, detail lost |
| 100 | 10 | 10 | Clearly noisy shadows |
| 1,000 | About 32 | About 32 | Moderate, visible grain |
| 10,000 | 100 | 100 | Clean midtones |
| 50,000 | About 224 | About 224 | Very clean highlights |
Since SNR grows with the square root of light, doubling the exposure improves SNR by about 1.41 times, and four times the light is needed to double it. This explains why one stop of extra exposure gives a noticeable but not dramatic improvement.
Shot noise compared with other noise sources
Shot noise is one of several noise components that add together in a raw file. Independent noise sources combine as the square root of the sum of their squares, so the largest component usually decides how the image looks.
| Noise source | Origin | Where it dominates | Can it be reduced |
|---|---|---|---|
| Photon shot noise | Random arrival of photons | Midtones and highlights | Only with more light |
| Read noise | Sensor readout electronics | Deep shadows, very low light | Better sensors, higher gain |
| Dark current shot noise | Thermal electrons in the pixel | Long exposures, warm sensors | Cooling, shorter exposures |
| Fixed pattern noise | Pixel-to-pixel differences | Shadows after heavy lifting | Calibration, dark frames |
| Quantization noise | Rounding in conversion | Rarely visible today | Higher bit depth |
The boundary between the regimes matters in practice. Modern sensors often have read noise of only a few electrons, so shot noise dominates almost everywhere except the darkest tones. In those shadows, read noise sets the floor and largely determines dynamic range.
Shot noise shows up as both luminance noise and chromatic noise. Each color channel behind the Bayer filter receives its own random count, so the variations do not match between red, green and blue pixels, which produces colored speckles after demosaicing.
What shot noise depends on
Shot noise depends on the total amount of light captured, not on the ISO setting. ISO only amplifies the signal after capture; the photons, and therefore their randomness, are already fixed by the exposure, the aperture and the sensor area.
| Factor | Effect on captured light | Effect on shot noise |
|---|---|---|
| Longer shutter time | More photons per pixel | Lower relative noise |
| Wider aperture | More light per area | Lower relative noise |
| Larger sensor at same framing | More total light for the image | Cleaner at equal output size |
| Higher ISO at fixed exposure | No change in photons | No change in shot noise |
| Downsampling the image | Combines several pixels | Lower noise per output pixel |
Pixel size is often misunderstood. A smaller pixel collects fewer photons and looks noisier at 100% view, but when images are compared at the same output size, the total light gathered by the sensor matters. Sensors of the same area and similar efficiency produce similar shot noise in the final image regardless of pixel count.
Quantum efficiency also plays a role. A sensor that converts a larger share of incoming photons into electrons collects a larger N from the same light, and therefore shows slightly lower shot noise at the same exposure.
Practical ways to minimize shot noise
The only direct remedy is collecting more photons. The table below summarizes techniques photographers use when noise matters.
| Strategy | How it helps | Trade-off |
|---|---|---|
| Expose to the right | Maximizes light without clipping | Risk of blown highlights |
| Use a faster lens | More light at the same shutter speed | Shallower depth of field |
| Stack several frames | Averages random variation | Needs a static scene |
| Add light to the scene | Raises signal directly | Changes the look of the light |
| Apply noise reduction | Smooths random variation | Loss of fine detail |
Frame stacking is especially effective because shot noise is random between exposures. Averaging four frames halves the noise, and sixteen frames reduce it to a quarter, which is why astrophotographers and smartphone night modes rely on many short captures.
Common misconceptions about shot noise
Many noise problems blamed on the camera are really shot noise from underexposure. The following points correct common misunderstandings.
- High ISO does not cause shot noise; the low light that required high ISO does.
- Better electronics cannot remove it, because it comes from the light itself.
- Brightening a dark file in editing reveals shot noise that was captured, and adds no new light.
- Absolute noise rises with exposure, yet the image looks cleaner because SNR improves.
- Shot noise is not a pattern; repeating patterns point to fixed pattern noise instead.
A non-technical view
Light arrives at a camera sensor as tiny particles, a bit like raindrops falling on a row of buckets. Even in steady rain, no two buckets collect exactly the same amount. Those small differences appear in a photo as grain, and that grain is shot noise.
In a heavy shower, the differences are tiny compared with the full buckets, so the result looks even. In a light drizzle, they are large in comparison. That is why photos taken in dim light look grainier, and why the real cure is simply more light.
A concrete example
A wildlife photographer shoots deer at dusk with a 400 mm f/5.6 lens. To freeze movement, she needs 1/500 s, and the camera meters ISO 12800. The frames show heavy grain in the fur and gray patches in the shadows, which is mostly shot noise from the small number of photons reaching each pixel.
The next evening, she switches to a 300 mm f/2.8 lens, which gathers two stops more light at the same shutter speed, and arrives twenty minutes earlier while more ambient light remains. The camera now meters ISO 3200, and she exposes slightly brighter without clipping the sky.
Because each pixel collects about four times more photons, the signal-to-noise ratio roughly doubles. The fur shows clean fine texture, and only light noise reduction is needed during processing.