# What Is RAW HDR+? Multi-Frame Merging Explained

> RAW HDR+ is a burst technique that merges several unprocessed RAW frames into one photo, cutting noise and recovering highlight and shadow detail.

Source: https://www.sensepose.com/blog/what-is-raw-hdr-plus
Updated: 2026-09-30

RAW HDR+ is a photo technique that merges a burst of unprocessed RAW frames
into one image with less noise and wider dynamic range. It is the mechanism
behind almost every "how did my phone do that?" moment of the last decade —
night mode, clean shadows, skies that aren't blown out. The name comes from
Google's HDR+ pipeline, first shipped on Nexus and Pixel phones, and the same
multi-frame approach underpins iPhone capture. It is the central idea in
[computational photography](/blog/what-is-computational-photography).

## What a burst is, and why RAW matters

Tap the shutter on a multi-frame camera and you aren't taking one photo. The
camera is already running a rolling buffer, and keeps the handful of frames
captured around your tap — typically 3 to 15, all at the *same* short exposure,
inside a fraction of a second.

**RAW** means those frames are the sensor's own readings: 10- to 14-bit linear
values with no white balance, tone curve, sharpening or noise reduction
applied. Linear is the operative word — a RAW value is proportional to the
photons that hit that pixel, so adding RAW frames together means something
physically. JPEG values have already been through a gamma curve, sharpening and
denoising; stacking those compounds whatever the first pass got wrong.

**Burst** means the frames are near-identical rather than bracketed, and this
is the counter-intuitive part. Classic HDR shoots one dark, one normal and one
bright frame and blends them. HDR+ instead shoots every frame *deliberately
underexposed*, short enough that even the bright parts stay below clipping,
then recovers the shadows after merging. Protect highlights in hardware,
because a clipped highlight is gone forever; fix shadows in software, where
merging can actually help.

## Why merging frames cuts noise

The grain in a dark photo is mostly *shot noise*: light arrives as discrete
photons, and the count landing in any pixel during a short exposure fluctuates
randomly. Fewer photons, noisier estimate.

The signal is the same in every frame of a burst; the noise is different, and
random. Average them and the signal stays put while the noise partially
cancels — the random component falls roughly with the square root of the frame
count. Four frames roughly halve it; sixteen roughly quarter it. That's why a
merged stack survives being lifted several stops in the shadows, and why night
modes spread light-gathering across many frames rather than one long,
blur-prone exposure — the trade-off behind
[night photography on Android](/blog/how-to-take-night-photos-android) and
[low-light camera apps](/blog/best-low-light-night-camera-app).

## How a merge recovers highlights and shadows

Dynamic range is the distance between the brightest and darkest detail a photo
can hold. A single short frame protects the sky but buries the shadows in
noise; a single long frame opens the shadows but clips the sky to flat white.

A merge escapes the choice. Because the stack is far cleaner than any of its
members, you can expose for the highlights and brighten the dark regions
afterwards, which the merged noise floor can absorb.

| What changes | Single RAW frame | RAW HDR+ merge |
|---|---|---|
| **Shadow noise** | Full sensor noise; limits how far you can lift | Falls roughly with √(frames); shadows survive lifting |
| **Highlights** | Must expose brighter, risking a clipped sky | Every frame underexposed, so highlights stay intact |
| **Dynamic range** | Bounded by one exposure | Wider, because clean shadows can be recovered |
| **Fine detail** | Whatever one frame caught | Sub-pixel shifts between frames add real detail back |
| **Moving subjects** | Frozen exactly as shot | Risk of ghosting; needs motion-aware merging |
| **Capture time** | One exposure | Longer — the burst must complete |

## The hard part: alignment and ghosting

Averaging frames is arithmetic. Deciding *which* pixels to average is the
engineering problem. Between frame one and frame eight, your hands moved — and
so did the leaves, the traffic, and the person you're photographing. Average
misaligned frames directly and you get a blurry, double-exposed mess, worse
than a single shot.

So a real pipeline does three things first. It **picks a reference frame**,
usually the sharpest of the burst. It **aligns the others in tiles** rather
than as whole images, so a shaken frame can be nudged into place region by
region at sub-pixel precision. And it **merges robustly**: where a tile
disagrees badly with the reference — a car moved, a branch swayed — that
frame's contribution is reduced or dropped.

That last step is ghost suppression, and it has an honest cost. A region where
motion was detected merges fewer frames, so it gets less noise reduction than
the static parts of the scene. Push too hard on a moving subject and you see
the failure directly: a faint second copy of a limb, a smeared hand.

## Why it works on a tiny phone sensor

A phone sensor is at a huge physical disadvantage — the reason a big camera
still wins on single-exposure quality, as covered in our
[phone camera vs DSLR](/blog/phone-camera-vs-dslr) comparison. You can't make
the sensor bigger. But light collected is area × time, and multi-frame capture
buys back the *time* axis: ten frames gather roughly ten frames' worth of light
with no single exposure long enough to blur.

That's the whole trick. A merge spends what phones have plenty of — compute and
a few hundred milliseconds — to buy what they lack, sensor area.

## What RAW HDR+ cannot do

- **Fast motion.** A burst takes time. Sports, kids and moving vehicles give
  the merger less to work with, and motion regions end up noisier.
- **Diminishing returns.** Noise falls with the square root of frame count, so
  4 frames to 16 costs 4× the capture time for half the noise — while the odds
  of something moving rise the whole way.
- **Clipped is clipped.** If a highlight is blown in *every* frame, no merge
  brings it back. Hence the deliberate underexposure.
- **It won't fix focus or bad light.** Merging improves noise, dynamic range
  and detail — not a missed focus or flat lighting.
- **Power and heat.** Aligning and merging RAW frames is real computation.

## Where SensePose fits

Most Android phones have a sensor perfectly capable of feeding a burst merge,
and a stock camera app that never does one. GCam ports are the usual
workaround, but they lock you to auto exposure and often wash out color on
non-Pixel hardware; ProShot and Open Camera give you sliders but no AI — a gap
covered in our [GCam vs AI camera app](/blog/gcam-vs-ai-camera-app) comparison.

**SensePose** takes a different route, and it's worth being clear: it does not
do RAW HDR+. It takes a single frame and enhances it on the phone as you shoot:
better colour and tone, cleaner low light from a mild on-device denoise, and
AI face enhance on arm64 phones with 3 GB RAM or more. It's point-and-shoot,
with a Natural or Punchy look, so it won't recover a highlight a single
exposure clipped.

The first 5 enhanced photos are free; after that, photos still save free as
the original at full resolution, and one small payment unlocks enhanced photos
forever, with no subscription and no watermark. There's no account or sign-in,
your photos never leave your phone, and enhancing works offline. It runs on
Android 8.0+. SensePose is [an AI camera app](/blog/best-ai-camera-app-android),
and it's on Google Play.

## FAQ

### What is RAW HDR+ in simple terms?

RAW HDR+ is a capture technique that shoots a rapid burst of unprocessed RAW
frames and merges them into one photo. Because the random noise differs in
every frame but the scene does not, averaging cancels much of the noise — which
lets the camera underexpose to protect bright highlights and brighten the
shadows afterwards.

### How is RAW HDR+ different from regular HDR?

Regular HDR brackets exposures: one dark, one normal and one bright frame,
blended. That handles contrast but does nothing for noise, and ghosts easily
because the frames differ so much. RAW HDR+ shoots many frames at the *same*
short exposure so highlights never clip, merges them to cut noise, then lifts
the shadows computationally — on linear RAW sensor data rather than
already-processed JPEGs.

### Does RAW HDR+ work in low light?

Yes, and low light is where it helps most. In dim scenes shot noise dominates,
and averaging frames is one of the few ways to reduce it without a long
exposure that would blur from hand shake. A burst gathers many frames' worth of
light while each frame stays short enough to be sharp. The limit is motion:
subject and photographer need to be reasonably still.

### Why do merged photos sometimes show ghosting?

Ghosting happens when something moves between frames and the merger blends the
mismatched positions together, leaving a faint second copy or a smeared edge.
Good pipelines detect this by comparing each frame tile against a reference
frame and dropping tiles that disagree. That suppresses the ghost, but those
regions merge fewer frames, so moving parts of a photo end up noisier than the
static background.

### Does RAW HDR+ need a flagship phone?

No. The technique needs RAW frame access through the camera API and enough
compute to align and merge a handful of frames — not a dedicated imaging chip.
Plenty of mid-range and budget phones can run a burst merge; most simply ship a
camera app that never attempts one.

---

Get SensePose on Google Play: https://play.google.com/store/apps/details?id=com.profazia.sensepose&referrer=utm_source%3Dsite%26utm_medium%3Dblog-what-is-raw-hdr-plus%26utm_campaign%3Dv2