Bank Security Camera Quality: What Makes Footage Usable


Bank camera quality is not about the highest megapixel number. It is about whether a still frame from the recording can identify a person: a face, a transaction, a license plate. That outcome comes from five things working together: camera placement, resolution, lighting, compression and the network that carries the video. A business that wants bank-grade footage needs a system designed around those five, not a camera picked from a spec sheet.
This guide is for owners and managers of offices, stores and other commercial sites who want footage that holds up when it matters. The standards banks follow are public, and the same logic applies to a cash wrap, a front desk or a server room.
TL;DR
Good footage means identifiable faces at every point where money, goods or people pass. It is a design result, not a resolution number.
Poor lighting is the most common reason video is degraded, so fix light and backlighting before upgrading cameras.
Place at least one camera at each transaction point so a face fills roughly 15% or more of the frame.
Compression and bandwidth can quietly destroy a good image. Size storage and network for the cameras you install.
A site assessment before purchase catches most of these problems while they are still cheap to fix.
Table of Contents
What bank-grade camera quality actually means
Banks are held to a simple purpose: the footage must help identify and prosecute. Federal Reserve regulation 12 CFR 208.61 requires member banks to keep a written security program that preserves evidence aiding identification and prosecution, and it lists a camera recording banking-office activity among the procedures a bank may use.
The regulation does not set a megapixel count. The technical guidance comes from forensic video groups. A forensic best-practices white paper says usable footage must capture uniquely identifying details such as facial features, tattoos, clothing patterns, license plates or held objects. It ties quality to placement, resolution, lighting, color, compression and frame rate.
The facial-comparison community says the same thing from the other direction. The FISWG image-factors document states that facial comparison is highly sensitive to image quality. Blur, compression artifacts or reduced resolution can reduce or eliminate the visibility of facial features.
So the test for any camera is practical: can you pause the recording at the moment someone stands at the counter and tell who they are?

Placement comes first
The SWGDE guidelines for commercial institutions call for at least one camera at each customer-transaction point: teller windows, ATMs, registers. The area of interest should cover about 15% or more of the field of view. In practice, that means a field of view about 3 feet wide for a 6-inch head, or about 6 feet wide for a 12-inch license plate.
The earlier SWGDE video security guidance adds a point many buyers miss. Identification requires seeing features like the eyes, ears, nose, mouth and chin, and overview cameras cannot be relied on for identification.
A bank branch therefore uses layers. The Axis bank case-study book describes overviews above tellers, the recycling area and vault, plus close-up face images at the entrance and exit. The forensic white paper likewise recommends overall, intermediate and close-up views.
For a business, translate that into three layers:
Overview: shows the whole room and how people move through it.
Context: shows a counter, door or stockroom with people at a usable size.
Identification: a face-height camera at the entrance, exit and each register or desk where cash or valuables change hands.
The New York State Department of Financial Services letter on preventing bank robberies recommends positioning cameras for full frontal images at teller stations and entrances. A camera mounted high and angled steeply records the top of a head. Mounting height and angle matter as much as the sensor.
If you are unsure where your gaps are, our guides on planning camera coverage and finding blind spots walk through the process.
Resolution: what you need and what is marketing
More pixels help only when they land on the subject. SWGDE sets a floor of 640x480 digital pixels, strongly recommends higher resolution, and says the recording resolution should match the camera’s resolution. A camera that captures in 4K but records at a lower setting delivers the lower setting.
The practical way to think about it is pixels on the face. A wide lens covering a whole lobby spreads its pixels over a large scene, so each face gets few of them. A narrower view at a counter gives the face far more. This is why one overview camera at 4K rarely replaces a dedicated camera at each counter.
Higher resolution does give you room to work. A camera maker’s executive quoted in ASMAG describes 4K as four times the resolution of HD, with much more digital zoom before pixelation, and suggests a bank might halve the cameras needed to cover tellers when moving from HD to 4K. Treat that as a design option to evaluate for your floor plan, not a rule.
Two questions settle most resolution decisions:
What must the camera show: a face, a plate, a receipt, or a general scene?
How far away, and how wide is the view?
Answer those and the resolution follows. Skip them and you end up paying for pixels that sit on an empty ceiling.
Lighting: the most common cause of bad footage
Both SWGDE documents state that poor lighting is the most common factor degrading video image quality. The 2015 guidance lists minimum illumination levels for bank environments:
Area | Minimum illumination |
Customer, office, hallway, stairway and exit areas | 275 to 333 lux |
Exterior ATM vestibules and drive-up lanes (24 hours) | 110 lux |
Exterior sidewalks, entrances, night depository areas | 55 lux |
Parking lots | 11 lux |
Lux is a measure of light falling on a surface. A phone light-meter app gives a rough idea, but the point is that a dim back office and a sunlit storefront are very different jobs for a camera.
Backlighting is the second problem. The SWGIT commercial-institution guidance warns that strong backlight or high-contrast scenes can shadow a suspect’s face and make identification difficult or impossible. It recommends considering non-infrared, high-dynamic-range cameras that work in low light. Anyone with a glass storefront or a door facing a bright window knows this scene: a camera aimed at an entrance records a silhouette at noon.

The Axis case-study book also names large windows and entrances as a key design consideration for branches.
Before blaming the camera, check whether the lighting changes through the day. Our articles on night vision problems and why cameras look blurry cover the common causes and fixes.
Frame rate, color and lens details
Frame rate decides whether you catch the moment. The forensic white paper says 30 frames per second is typically sufficient for smooth normal motion, and lower rates risk missing key moments. A hand passing over a counter in a fraction of a second can fall between frames at a low rate.
FISWG’s list of image factors is a good checklist for what else can go wrong: artifacts, distortion, lighting, noise, obstructions, pose, sharpness and resolution. It also notes that blooming, where sensor over-saturation causes chromatic aberration and lens flare, can hurt facial detail. In plain terms, a dirty dome, a spider web on the lens or a sign hanging in front of the view can ruin a recording as surely as a cheap sensor.
Compression, storage and the network
This is where a good camera becomes bad footage. IP cameras send video across your network as compressed digital data. The DHS Video Security Systems Technology handbook (February 2025) says this places high demand on bandwidth, with image quality as the tradeoff. It recommends separating the video stream from the primary network. It also warns that if compression is too aggressive and too many pixels are removed, faces or objects may not be recognizable.
The same handbook notes that H.265, the most commonly used current standard, needs roughly half the bandwidth of other standards while providing better quality. It also shows why compression is unavoidable: 2 hours of uncompressed video from one camera at 30 images per second and 1 MB per image equals 216 GB.
Resolution multiplies the load. EtherWAN’s network guide explains that at a constant bitrate, raising resolution lowers image quality, so a 4K stream needs about four times the bitrate of 1080p to hold the same quality. A US surveillance storage report (April 2026) puts a continuously recording 4K camera at roughly 300 to 400 GB a day in H.264, against 70 to 90 GB for 1080p. It also notes that financial-institution regulations are driving longer retention periods.
A worked example shows the point. Take 20 cameras on a small office and retail site, each streaming 4 Mbps. That is 80 Mbps of video before network overhead. If those cameras share a switch with point-of-sale traffic and office WiFi, a busy afternoon can cause dropped frames in the recording and slow card payments at the same time. A dedicated video network, correctly cabled and sized, avoids both.
Three decisions to make before you buy:
Retention: how many days of footage you must keep. Multiply daily storage per camera by camera count and days.
Recorder type: network or digital recorder, matched to camera type. Our NVR and DVR comparison explains the choice, and the analog vs. IP cameras guide covers the camera side.
Cabling: every IP camera needs a clean, tested cable run back to the switch. Weak cabling shows up as intermittent dropouts that are hard to trace later.
Why this is a system job, not a product purchase
None of the factors above lives inside the camera box. Placement depends on your floor plan. Lighting depends on your windows and fixtures. Storage depends on your retention needs. Bandwidth depends on your network. Buying cameras online and mounting them at convenient spots tends to produce a system that records everything and identifies no one.
The work is the sequence: walk the site, mark the transaction points and entrances, measure the light, choose lenses for each target, plan cable paths and switch capacity, then size recording. Our overview of what matters most in business security cameras and our guide to CCTV installation that works cover that sequence in more detail.
Promo: bank-style planning for your site
Tekmatik installs CCTV and security camera systems for offices and retail locations, and we also handle the structured cabling and network setup those cameras depend on. Because one team covers both the cameras and the network behind them, the questions about placement, cable runs and bandwidth are planned together. Our installations are carried out with attention to minimizing disruption and leaving the space clean.
We start with an on-site assessment, a one-off $99 fee, after an initial phone consultation. You can request a consultation to book it, or see our CCTV and security services.

If you manage several locations or want ongoing support after installation, see our Enterprise Plans.
Perspective
Most of the blurry or useless footage we hear about traces back to the design, not the camera. A face-height camera at the door, even light on the counter, a clean cable run and enough network capacity do more for identification than a bigger megapixel number. That is why we treat CCTV as part of the whole technology infrastructure of a business. Cameras, cabling, network and recording all have to be planned together, and we are set up to do that because we install and support each of them. Our approach is to walk the site first, decide what each camera must see, and then install precisely and cleanly so your operations keep running. After installation, our maintenance and troubleshooting services keep the system recording what you expect it to. Tekmatik
FAQ
Is 4K always better for a bank security camera?
Not automatically. 4K gives about four times the detail of HD and more room to zoom, but it also needs about four times the bitrate to hold the same quality, per EtherWAN. It pays off where it is paired with proper placement, light and network capacity.
How many cameras does a bank branch or similar business need?
The guidelines call for at least one camera at every transaction point, plus overview and entrance and exit views. The count comes from your floor plan: each register, desk, door and cash-handling area needs its own identification-level view.
How long should footage be kept?
For banks, retention requirements are set by financial regulators and are being extended, according to the April 2026 storage report. For a business, decide on a period that matches how long incidents or disputes typically surface, then size storage for it. If you handle regulated data, ask your attorney or compliance adviser which rules apply.
Do cameras need extra lighting?
Often yes. Poor lighting is the most common cause of degraded video, and the 2015 SWGDE guidance sets levels such as 275 to 333 lux in customer areas. An on-site assessment can show where light is uneven or where backlighting from windows will shadow faces.
Sources
Recommended
Comments