The footage is perfectly sharp. The doorway is bright and clean, you can read the lettering on the glass, you can see the parked cars outside. And the person walking in is a black outline with no face in it.
This is the most common disappointment in commercial video, and it is remarkably consistent. It happens at glass storefronts, lobby vestibules, roller doors and parkade entries, on cheap cameras and expensive ones, and it happens again after the upgrade that was sold to fix it.
It survives an upgrade because it is not a camera-quality problem. It is a geometry problem. The light is behind the person you care about, and the camera is looking straight into it.
One scene, one exposure
A camera has to choose a single exposure for the whole frame. That is the entire root of it.
Exposure is how long the sensor collects light, how wide the lens aperture is, and how much the signal is amplified afterwards. Set it for the bright doorway and everything darker goes to black. Set it for the person and the doorway blows out to a featureless white rectangle — and the person, lit only by what spills back off the floor and walls, is still barely there.
The camera makes that choice by measuring the scene and picking a middle. In a backlit entrance the bright area is usually the larger part of the frame, so the middle sits close to the doorway's brightness. The face, a small fraction of the pixels, gets whatever is left. Black outline.
Nothing about this improves with resolution. A silhouette rendered at four times the pixel count is a higher-resolution silhouette. That is worth saying plainly, because "more megapixels" is the reflexive answer to almost every complaint about footage and here it is simply the wrong axis. The pixels are not missing. The tonal information inside them is.
Dynamic range is the actual quantity
The relevant property is dynamic range: the span between the brightest and the darkest thing a sensor can record in one frame while keeping usable detail at both ends.
Every scene has a range of brightness, and every sensor has a range it can capture. When the scene's range is wider, something is discarded — highlights clip to white, shadows crush to black, or both. A backlit entrance is one of the widest-range scenes a building routinely contains: direct daylight through glass against an interior lit for human comfort. It gets worse, not better, on a bright day.
One thing follows, and it matters more than any product choice: the problem is the ratio between the bright part and the dark part, not the absolute brightness of either. A dim interior makes it worse, and so does a brighter exterior. Anything that narrows the ratio helps. That is the whole lever.
What WDR actually does
Wide dynamic range — WDR, sometimes sold as HDR — is the manufacturer's answer to the ratio problem, and it does real work.
The standard implementation captures the scene more than once in quick succession at different exposures: a short one that holds the bright doorway without clipping, a longer one that lifts the person out of the shadow. The camera combines them, taking highlight detail from the short exposure and shadow detail from the long one. Some sensors do it by reading different pixels at different sensitivities rather than in sequence.
It works. On a well-implemented camera in a backlit doorway the difference between WDR off and WDR on is not subtle. But its limits are exactly where entrance cameras live:
Motion. The captures are not simultaneous. A person walking briskly through the door is in a slightly different place in each one, so the merge reconciles images that disagree. The visible result is ghosting, doubled edges or a smeared halo around the moving subject — often around the head, which is the part you needed. An entrance is a place where everything is moving.
Noise. The long exposure that rescues the shadows amplifies whatever noise is in them. Lifting a face out of deep shadow can hand you a face made of grain.
Frame rate. Combining multiple captures costs processing time. On some cameras the strongest WDR mode reduces the maximum frame rate, trading one kind of usable footage for another.
Physics. WDR extends the range a camera can cope with. It does not make it unlimited. Point a camera at the sun through a glass door and no processing mode saves it.
WDR is a compensation, and a good one. It is still compensating for a decision made when someone picked where to put the bracket.
True WDR and digital WDR are not the same product
This distinction is where a lot of money gets spent badly, because both appear on a spec sheet as the letters WDR.
True WDR — also called multi-exposure, sensor or hardware WDR — is what is described above. The sensor genuinely captures multiple exposures and merges them. New information is created, because the camera looked at the scene more than once.
Digital WDR — DWDR, digital tone mapping, or backlight compensation depending on the vendor — is applied to a single captured frame. Software brightens the dark regions and pulls down the bright ones. No additional capture happens.
The difference is decisive: digital WDR cannot recover information that was never recorded. If the face landed in a range the sensor rendered as near-black, lifting it produces a brighter, noisier near-black. If the doorway clipped to pure white, no tone mapping restores what was outside the glass. Digital processing redistributes what is there. It does not add.
Digital WDR is not useless. On moderate contrast it improves an image, and because it runs on a single frame it introduces none of the ghosting that merging can cause. But on a severely backlit entrance it is close to cosmetic.
The practical consequence is a question rather than a rule. When a proposal says WDR, ask whether it is multi-exposure at the sensor or digital processing on a single frame. It has a definite answer, and a supplier who works with these cameras daily will give it without hesitation.
The dB trap
Now the part that makes spec-sheet shopping unreliable.
WDR performance is usually published as a figure in decibels. Decibels are a logarithmic measure of ratio, so a larger number does describe a wider range — the direction of the relationship is real. The trap is the assumption that follows: that a larger published number from one vendor beats a smaller one from another. It frequently does not, for reasons that have nothing to do with dishonesty:
The test scene is chosen by the vendor. A dynamic range figure is measured against a target under lighting conditions someone selected. Manufacturers select differently. Two figures from two setups are two different measurements wearing the same unit.
The threshold for "usable" is a judgement. Where detail stops being detail and becomes noise is a defined criterion in a lab and a matter of taste in a brochure. Move that threshold slightly and the published number moves substantially, because the scale is logarithmic.
The figure may describe the sensor, not the camera. A sensor's raw capability, quoted from component data, is not what survives the lens, the processing pipeline and the compression that follows. Recovered shadow detail that a bit-rate-constrained encoder discards on the way to the recorder is not detail you have.
So a bigger dB number, measured under conditions the vendor chose, is not comparable across vendors. It is not a lie and it is not meaningless — within one manufacturer's range, tested consistently, it will usually rank their own models correctly. Across brands it is an apples-to-oranges number in a units suit.
We are deliberately not printing example figures here, because doing so would undercut the point. The honest test is not a number: it is footage of a real person, at your door, at the worst time of day, through the camera being proposed.
The fixes, in priority order
Backlight fixes have a clear ranking, and almost everyone tries them in reverse.
1. Move the camera off the axis of the light
The first and best fix costs nothing but thought. If the camera is not looking into the bright source, the ratio inside the frame collapses.
In practice: mount to the side of the doorway looking across the threshold rather than opposite it looking out; or mount inside facing in, so arriving people are captured after they have entered and turned, with the bright glass behind the camera rather than behind them; or angle down and across a lobby so the frame fills with interior surfaces rather than window.
This is a trade, and worth naming as one. A camera facing into the room sees a face well but sees less of the approach outside. Many entrances end up wanting two cameras with two jobs — one wide view for context, one tight off-axis view for faces — rather than one camera failing at both.
Sun angle deserves a note, because it is seasonal and people plan around the day they did the site walk. A doorway that is fine at midday in summer can be unusable on a winter afternoon, when the sun sits low and shines directly down the axis of the lens. Axis notes the hazard in its own product documentation for plate reading, where "direct sunlight can distort the image, for example, during sunrise and sunset". The same geometry applies to faces at your front door. Check which way the door faces, and think about winter, before the bracket goes up.
2. Add light on the subject side
If the camera cannot move, change the ratio the other way: put light on the person.
This is the least glamorous fix and often the most effective. Well-placed, even illumination on the inside face of the entrance — the side of the person the camera can see — narrows the gap between subject and background directly. It is frequently a lighting change the building wanted anyway.
Two cautions. Light aimed at the camera creates a new bright source and a new problem, so aim it at the subject, not the lens. And infrared illumination, which many cameras carry built in, does not help here: it is a night-time tool and does nothing about daytime backlight. Thermal imaging is likewise not an answer — Axis is explicit that its thermal cameras are a detection technology and that thermal imaging "doesn't allow for clear identification", which is precisely what a silhouette already fails at.
3. Then use WDR
With the geometry improved and the subject lit, WDR closes the remaining gap, which is what it is good at. Applied to a scene aimed into a sunlit glass wall with an unlit interior behind it, it is being asked to substitute for a decision nobody made.
Ordering the fixes this way also changes what an upgrade costs. Reposition and add light and you may keep the cameras you own. Skip both and buy the highest-specified WDR camera on the market, and you will have bought a very capable camera and pointed it at the same impossible scene.
What to ask
Five questions, answerable by anyone who does this work properly:
- Which way does this door face, and what does the frame look like at the worst hour and worst season? Not at the hour of the site visit.
- Can this camera move off the axis of the light, and what do we lose if it does? There is usually a trade; you want it named.
- Is the WDR multi-exposure at the sensor, or digital processing on a single frame?
- What is the ghosting behaviour on a walking subject at the strongest WDR setting? This failure mode shows up only after installation.
- Can you show me footage of a person entering this door, on this camera, at the worst time of day? Everything above is a proxy for this question.
The entrance is the hardest lighting problem in most buildings and it is usually treated as the easiest — one camera, over the door, pointing out, because that is where a camera obviously goes. It is the one position guaranteed to put the brightest thing in the scene directly behind the one face you actually needed.
