Glossary
Computational photography
Techniques that combine multiple frames, sensor data, and machine learning to produce images beyond what a single exposure could capture. The reason a 1/2.5-inch phone sensor can rival a full-frame camera on social media.
Computational photography is the software half of modern smartphone cameras. Hardware (sensor, lens) captures input; the SoC does the heavy lifting.
Key techniques
- HDR+ / Smart HDR / Deep Fusion. Capture multiple exposures, merge for shadow + highlight detail.
- Night Mode. Long-exposure capture aligned and stacked across many frames.
- Portrait mode. Depth estimation (lidar, dual cameras, or ML) + synthetic background blur.
- Pixel binning. Combine 2×2 or 3×3 sensor pixels into one for low light.
- ML denoise. Trained networks remove noise while preserving texture.
Tradeoffs
Computational pipelines occasionally produce artifacts: over-sharpened text, plastic skin, ghosting at scene boundaries. RAW capture (Apple ProRAW, Google Pixel RAW) preserves pipeline output but gives the editor recovery flexibility.
Where this matters
Categories that use computational photography
Continue reading