How Does Image Noise Occur? The Effects of High ISO, Long Exposures, and Underexposure

Camera noise comes from photon shot noise, read noise, and thermal noise. This guide distinguishes high ISO, long exposures, and lifting underexposed images, then gives practical shooting and post-processing noise-reduction strategies.

CameraLore Editorial Team Updated August 17, 2026

Image noise comes from several sources, including statistical fluctuations in photon arrival, sensor noise, and readout-circuit noise. High ISO often makes existing noise more visible; long exposures also increase dark current and hot pixels. Lifting a severely underexposed image amplifies both its weak signal and its noise.

Why Isn’t Noise a Single Phenomenon?

Luminance noise appears as variations in grain brightness, while color noise appears as random color speckles. Fixed-pattern noise can form stripes or repeating structures, and hot pixels appear as fixed bright points during long exposures. The EMVA 1288 Release 4.0 Linear standard measures temporal dark noise, system gain, and quantum efficiency separately, showing that reproducible noise comparisons require a defined method rather than a glance at one thumbnail. JPEG noise reduction, sharpening, and compression all change the appearance, so RAW files are the best place to investigate the source, ideally by comparing multiple images made under the same conditions.

Sensor dust stays in a fixed position but usually appears as a dark spot, especially against an even background at a small aperture. It is not random noise. Do not use noise-reduction software to deal with dust that should be cleaned instead.

Where Do Photon Noise and Read Noise Come From?

Photons reach each pixel as a random process, so even a uniform light source produces statistical variation. The fewer photons there are, the greater that variation is in relative terms. The Photon Transfer Curve Primer describes the relationships among photon shot noise, read noise, and signal level separately. Shot noise cannot be eliminated completely with more expensive circuitry; increasing the actual exposure is the fundamental way to improve it. Read noise comes from the charge-conversion, amplification, and analog-to-digital conversion chain. It can be reduced, but it does not automatically become zero.

Comparisons between sensor formats and pixel counts should use the same composition, depth of field, shutter speed, and output size. Comparing a single pixel only at 100% magnification exaggerates the effect of pixel-density differences; see Do More Megapixels Mean Better Image Quality? for the relevant boundary.

What Does High ISO Really Do?

ISO primarily changes recording gain and output brightness; it does not make pixels receive more light. High ISO is often used in low light, so the sample images already contain relatively little photon signal. The gain then makes the noise more visible. On some cameras, increasing analog gain can reduce the relative impact of subsequent read noise, but it comes at the cost of highlight headroom.

The right approach is not to use the lowest ISO at all times. First make sure the aperture and shutter speed work for the image, then choose an appropriate ISO. See How to Set Camera ISO for the complete procedure.

Why Does Thermal Noise Appear During Long Exposures?

As the sensor warms and exposure times grow longer, dark current accumulates. This can produce colored bright pixels, glow, and fixed-pattern artifacts. In-camera long-exposure noise reduction typically makes a dark frame of the same length after the actual exposure and subtracts it. This can reduce fixed hot pixels, but it keeps the camera occupied for roughly twice as long and cannot eliminate random photon noise.

For astrophotography sequences or stacking, you can shoot dark frames and process them with a dedicated workflow. Do not assume that subtraction in every software package is suitable for images with movement, temperature changes, or in-camera correction already applied.

Why Does Lifting an Underexposed Image Look Dirtier?

Underexposure means there is less useful signal. Raising exposure by two stops in post-processing amplifies the signal, photon fluctuations, and read noise together. If JPEG processing has already darkened the shadows, tonal and compression losses will also be more apparent. As long as important highlights are not saturated, opening the aperture or using a slower shutter speed to capture more light is usually more effective than brightening the image afterward.

High-contrast scenes require balancing highlight capacity; do not sacrifice important highlights just to keep the shadows clean. Dynamic Range and RAW Latitude explains this boundary.

How Can You Reduce Noise During Capture and Post-Processing?

Increase the actual exposure as long as doing so does not create motion blur or depth-of-field problems.
Avoid unnecessary extended ISO settings and keep the RAW file for post-processing.
Control the camera’s temperature during long exposures; use dark frames or in-camera long-exposure noise reduction when necessary.
For multiple static frames, align and average them. Random noise will decrease, but moving areas may produce artifacts.
In post-processing, correct the lens and exposure first, then balance luminance noise reduction, color noise reduction, and detail. Sharpen for the final output size at the end.

Noise reduction necessarily involves a judgment between random texture and genuine detail. Skin, feathers, and grass are easily smoothed away, so evaluate the result at the final output size rather than by zooming in indefinitely.