Open the camera on a Google Pixel and press the shutter. Behind that single tap, the phone is doing something most cameras do not: it silently captures a rapid burst of underexposed frames, aligns them, merges them, and re-renders the result into one image with far more dynamic range than the sensor could physically capture in a single shot. That process is HDR+, and it is one of the clearest demonstrations of what “computational photography” actually means.

The Problem HDR+ Solves

Smartphone cameras are physically small. A tiny lens gathers little light, and tiny sensor pixels have limited dynamic range — the span between the brightest highlight that does not blow out to white and the darkest shadow that does not collapse to black. In dim scenes the result is noise; in high-contrast scenes, such as a person backlit by a bright sky, a conventional single shot must choose between a blown-out sky and a subject reduced to a silhouette.

Traditional HDR techniques try to fix this with exposure bracketing: capture one bright, one mid, and one dark exposure, then blend them. But bracketing has a cost. The long exposure blows out highlights, the short exposure is noisy, and because every frame looks different, aligning them is hard — which is why bracketed HDR so often produces ghosts and halos around moving objects.

Burst Photography: The Core Idea

Google’s key insight, first shipped in the Google Camera app on the Nexus 5 and Nexus 6 and refined ever since, was to skip bracketing entirely. Instead, HDR+ captures a rapid burst of full-resolution raw frames — typically three to fifteen — all deliberately underexposed and all at the same short exposure time.

Underexposing is what saves the highlights. By keeping every exposure short, HDR+ ensures the bright sky never clips to white in the first place. The trade-off is that shadows come out noisy — but HDR+ fixes that by shooting many frames and averaging them together, which drives the random noise down. As Google Research describes it, the capture strategy is based on underexposure, and the noise penalty is offset through the use of burst photography.

Alignment and “Lucky Imaging”

Before frames can be averaged, they must be lined up. Hand shake and moving subjects mean no two frames are pixel-identical. HDR+ begins by picking the sharpest single frame in the burst as a reference — a technique astronomers call “lucky imaging,” where you keep the best moment from a stream of noisy captures. The remaining frames are then algorithmically aligned to that reference.

Because every frame uses the same exposure, they all look similar, which makes this alignment robust. That is the direct payoff of abandoning bracketing: HDR+ images are essentially free of the ghosting and double-image artifacts common in other HDR software.

Merging: Averaging Away the Noise

Once aligned, the burst is merged into a single computational raw image. In the original pipeline this is a tile-based temporal denoising step: each frame is divided into tiles, corresponding tiles across the burst are stacked, and a Wiener-filter-style combination in the frequency domain blends them into one low-noise tile. The merged result holds reduced noise and increased dynamic range while retaining the sharpness of the chosen reference frame.

The output of this merge is a 14-bit intermediate “linear RGB” image whose pixel values are proportional to scene brightness — effectively a raw photo with far more dynamic range than any single sensor capture could provide.

Tone Mapping: The Signature HDR+ Look

That 14-bit image cannot be displayed directly; a screen needs 8-bit output. Squashing the range down is called tone mapping, and it is what gives HDR+ photographs their distinctive look. Instead of applying one global curve — which cannot preserve detail in both deep shadows and bright highlights at once — HDR+ uses a local tone mapping algorithm that effectively applies a different curve to different regions, depending on local brightness, texture, and noise.

The result is an image with detail everywhere: a blue sky, a properly exposed face, and recoverable shadow detail, all in one shot with natural-looking edges.

From Nexus to Pixel: How HDR+ Evolved

HDR+ did not stand still. Zero shutter lag let later Pixel phones capture the burst even before you press the button, reusing the frames already streaming through the viewfinder. Dedicated silicon — the Pixel Visual Core on the Pixel 2 and 3, and the Pixel Neural Core on the Pixel 4 — accelerated the processing and opened it to third-party apps. HDR+ with Bracketing added a single long-exposure frame after the shutter press for scenes with extreme dynamic range, paired with a per-pixel spatial merge that prevents ghosting. Night Sight, introduced with the Pixel 3, reused the same burst-alignment-and-merge machinery with longer exposures for near-dark scenes, and Live HDR+ on the Pixel 4 used a machine-learning approximation (HDRnet) to preview the final look in real time.

Each of those features is a branch of the same tree, but the trunk has never changed: shoot a burst, align it, merge it, and tone-map it into one better-than-possible photograph.

Conclusion

HDR+ is the clearest proof that a modern phone camera is as much a software product as a hardware one. By replacing a single exposure with an aligned and averaged burst, Google turned a tiny, noisy sensor into one that outshoots its physical limits — no tripod, no manual bracketing, no editing required. The next time you tap the shutter on a Pixel, remember: the camera is computing your photo as much as it is taking it.

FAQ

Does HDR+ use exposure bracketing?

No. Classic HDR+ deliberately avoids bracketing. Every frame in the burst uses the same short exposure, which protects highlights and keeps alignment robust. A later feature, “HDR+ with Bracketing,” does add one long-exposure frame for extreme scenes, but the core HDR+ pipeline does not.

How many photos does HDR+ capture?

Depending on the scene and lighting, HDR+ captures roughly three to fifteen underexposed frames, plus — in bracketing mode — one long exposure captured after the shutter press.

Why don’t HDR+ photos show ghosting?

Because all burst frames share the same exposure, they look alike and align cleanly, and HDR+ merges them with a per-pixel algorithm that decides which content to blend. This avoids the ghosts and double images seen in bracketed HDR.