How does computational photography enhance low-light shots on modern flagship phones?

08 Ağu 14:52

Modern flagship devices typically use a three-stage process: a burst of short-exposure frames (often 3–7 shots), an on-chip HDR merge, and a final AI-driven denoiser. The sensor’s readout speed allows the ISP to capture several 1/30s exposures while the scene remains static enough for alignment. These raw frames are first combined using a tone-mapping HDR algorithm that extracts detail from the brightest and darkest captures, effectively expanding the dynamic range before any pixel-level cleanup. After the HDR blend, a neural network model—trained on millions of low-light samples—predicts and reduces residual shot noise while also sharpening edges that often blur during the merging process. If you compare this to older single-frame pipelines, like the Galaxy S20’s “Night mode” (which relied mainly on a long exposure and basic denoising), the newer multi-frame approach in the Pixel 8 or Xperia 1 IV delivers noticeably better signal-to-noise ratio and more natural colors, especially when your hand isn’t perfectly steady. The trade-off is added processing latency (typically 0.8–1.2 seconds from shutter to preview) and a slight battery drain—around a 5–8% increase per shot because the ISP has to power the sensor at higher frame rates and run the AI accelerator. Some manufacturers address this by limiting multi-frame mode to a few seconds of exposure or downgrading the AI denoiser to a lower-precision model when the battery is low. So the choice is clear: better low-light quality comes at the cost of a small delay and extra power usage, which is why many phones let you toggle “Night mode” on or off based on whether you prioritize speed or quality.