M12 Lenses are miniature lenses for embedded cameras, computer vision, and human viewing...
Rectilinear S-mount lenses that keep straight lines straight, from 1.8mm wide angle to 35mm telephoto. Datasheets with distortion plots are on each product page.
Inspection systems, mobile robots, and ADAS cameras read geometry straight off the sensor, so the distortion budget gets set before anything else. The notes below cover the field of view tradeoff, what software correction costs, and which designs hold the straightest lines.
25 rectilinear and near-rectilinear S-mount designs from $19. Titles state the nominal focal length; datasheets carry the distortion plot. Sold-out designs are hidden; contact engineering for restock dates.
No distortion means very low residual distortion, not zero. An ideal design can be exactly rectilinear on paper; what rules out a literal zero is manufacturing, since surface figure, centering, spacing, and material tolerances leave a small residual on every real unit. A good rectilinear design holds that residual small enough that straight lines stay straight for measurement and machine vision work.
Rectilinear projection maps straight scene lines to straight image lines under a central projection. That is a narrower promise than keeping geometry intact: perspective stays, so lengths, angles, areas, and magnification still change with depth and position in the field. What it buys is that measurement code and detection models see the projection they already assume, without a dewarp step.
Use these where residual distortion eats the error budget:
For the distortion mechanics behind these claims, see wide-angle lenses and fisheye camera lens distortion.
Low distortion M12 lenses accept a small residual distortion by design. Straight lines still read straight and calibration stays simple; coverage takes priority over metrology-grade accuracy.
Use these when the camera navigates rather than measures:
Distortion is driven by field angle, not by focal length on its own, so wider coverage on the same sensor is harder to keep straight: a 6mm design usually holds cleaner geometry than a 2.8mm there, compared at the same image height under the same convention. Read the curve rather than the corner number alone, since a mustache profile changes sign across the field and can peak short of the edge. Fighting barrel distortion? First ask whether you need that wide a view; narrowing the field sidesteps the problem.
You can dewarp in software, at a cost that depends on the correction. Resampling a whole image adds latency and redistributes local sampling: barrel correction stretches the outer field and lowers effective resolution there, while the output projection and the field you ask for decide whether anything gets cropped. Pipelines that only need geometry often apply the model to feature coordinates instead and never resample the imagery at all. For real-time systems or pixel-level measurement, a lens that needs little correction is still the cheaper starting point.
A lens specified for a 1/2" image circle does not cover a 1" sensor. What happens outside the rated circle depends on how that boundary was defined: an illumination limit gives falloff and then a hard mechanical cutoff, while an MTF-defined limit can stay lit but drop below the resolution the spec promises. Distortion beyond the rated field is unspecified rather than automatically worse. Verify active-area dimensions against CMOS sensor sizes and check chief ray angle compatibility with your sensor.
Then verify coverage and framing with the camera field of view calculator.
| Feature | No Distortion | Low Distortion | Fisheye |
|---|---|---|---|
| Projection type | Rectilinear | Near-rectilinear | Curvilinear |
| Geometry accuracy | Highest | High | Low |
| Measurement suitability | Yes | Depends on tolerance | No |
| Typical field of view | Narrow to moderate | Moderate to wide | Ultra-wide, 180° and beyond |
Need the ultra-wide end instead? Browse M12 fisheye lenses, where the curvature is the projection working as designed.
Distortion directly limits measurement accuracy: uncorrected barrel stretches dimensions toward the frame edge, and on tight-tolerance parts that error turns a pass into a fail. See our guide on image quality and computer vision.
Navigation stacks calibrate more simply on near-rectilinear input. A low distortion wide angle lens gives mobile robots scene coverage without pushing the pipeline into fisheye calibration models. Platform picks are in lenses for robotics and lenses for drones.
Lane detection works better when lanes render straight, and neural nets trained on rectilinear data lose accuracy behind curved geometry. See lenses for autonomous vehicles and our IP67-rated automotive M12 lenses.
Pick-and-place and alignment systems need a consistent pixel-to-millimeter mapping. Stable distortion is not the enemy there, since calibration absorbs it. The trouble starts when the mapping moves after calibration: someone refocuses, the lens shifts in its thread, temperature changes the assembly, or a replacement unit differs from the one that was calibrated. Then the arm lands 2mm off target and someone is debugging at 2am.
Spec the lens before you buy:
Lens distortion is the departure of image height from the lens’s stated projection: magnification changes with field angle, so straight lines render curved even in a sharply focused image. It is a geometric aberration, separate from blur or vignetting. In a third-order-dominated design the magnitude rises with field height, but real profiles are not always monotonic, and a mustache profile runs barrel near the center and pincushion toward the corners. Distortion is quoted against a reference mapping and a measurement image circle, which is why the same lens can carry different numbers on different sensors.
Barrel distortion bows straight lines outward and is common in wide angle lenses. Pincushion distortion pinches lines inward and shows up in telephoto designs. Mustache distortion mixes the two: barrel near center, pincushion at the corners. Sign convention: negative means barrel, positive means pincushion.
Barrel distortion is where magnification decreases with distance from the optical axis, so straight lines curve outward like a barrel. For conventional lenses it is a third-order (Seidel) aberration that grows with field angle. It also buys coverage: at the same focal length and the same image height on the same sensor, a design that accepts more barrel distortion maps a wider field angle into that image height, which is why wide angle and fisheye designs carry the most.
A rectilinear lens maps ray angle to image height as r = f·tan(θ), which keeps straight lines in the scene straight in the image. Every lens in this collection is rectilinear or near-rectilinear; fisheye lenses intentionally use different projections to reach 180° and beyond. Rectilinear rendering is what measurement, inspection, and lane detection algorithms assume.
In this collection, the No Distortion 4.4mm (CIL043) carries the tightest label, and the telephoto end tends to measure low: a longer focal length covers a narrower field on the same sensor, which is easier to correct, so a 6mm or 8mm design typically measures lower than a 2.8mm under the same convention and image height. Focal length does not set distortion by itself; field angle, sensor format, and the optical design do, and a mustache profile can peak short of the corner. Check the distortion plot on the datasheet rather than the product name; the measurement image circle matters.
A marketing term. A design can be exactly rectilinear on paper, but no production lens measures exactly zero, because fabrication and assembly tolerances leave a residual on every unit. A good rectilinear design holds that residual low enough that straight lines look straight and the error disappears inside your measurement budget. Commonlands lists distortion per lens on the datasheet, so you can hold the claim against a number.
Yes, within limits. Calibration models the distortion (Brown-Conrady for near-rectilinear lenses, Kannala-Brandt for fisheye). What it costs depends on how the model gets applied. Dewarping the whole image adds latency and resamples it, so local sampling and sharpness change across the field, and the output projection and the field you request decide whether the frame gets cropped. Correcting feature coordinates against the same model leaves the pixels untouched. For measurement and latency-sensitive pipelines, starting with clean optics beats fixing geometry every frame.
Start with the optics. Pick a rectilinear design with low residual distortion, and do not buy more field of view than the application needs, since a wider field angle is harder to correct. Match the lens image circle to your sensor so the corners stay inside the well-corrected field. Then check coverage with the FOV calculator, which accounts for distortion.
Tell us your sensor format and field of view requirements, and we’ll point you to the right lens. Samples ship the same day for orders placed before 12 PM PT. No minimum order.
