What is computational photography?
Using processing rather than optics to produce an image — capturing multiple frames and combining them computationally, which is how a phone with a tiny lens produces photographs a small sensor should not be capable of.
The core technique: multi-frame capture. When you press the shutter on a modern phone, it is not taking one photograph. It has been continuously capturing frames in a buffer, and it selects and merges several — aligning them, combining the best information from each, and discarding the rest.
What this enables:
Noise reduction. Noise is random; the signal is not. Averaging several frames cancels much of the noise while preserving detail, which is why phone night photography improved so dramatically.
HDR. Combining frames exposed differently retains detail in bright and dark areas simultaneously — beyond what the sensor can capture in a single exposure.
Night modes, which extend capture over seconds while compensating for hand movement, producing results handheld that would previously have required a tripod.
Portrait mode. Depth is estimated — from multiple lenses, a dedicated depth sensor, or machine learning on a single image — and background blur is applied artificially. The characteristic errors around hair and glasses reveal that it is a segmentation problem rather than optics.
Super-resolution, using tiny hand movements between frames to reconstruct detail beyond the sensor's nominal resolution.
Semantic processing — the device identifies faces, sky, skin and foliage and processes each differently. This is why skies are more saturated than they were and skin is smoothed.
The questions it raises. At what point does an image stop being a photograph? Devices have been found substituting generated detail — notably in images of the moon — and the line between enhancement and fabrication is genuinely unclear.
Why photographs from different phones look so different despite similar hardware: the processing is the product, and each manufacturer has a distinct interpretation of what a good image looks like.
Shooting RAW bypasses most of it, which is why RAW files from phones frequently look worse.