Camera calibration software
Online camera calibration software, from photos to parameters.
Print a pattern, take 15–30 photos and upload them. CalibrX detects the board, solves pinhole, wide-angle and fisheye models and gives you the camera matrix, distortion coefficients and files for OpenCV, ROS and Python — in your browser, with no code.
Quick answer
What is camera calibration software?
Camera calibration software estimates a camera's intrinsic parameters and lens distortion from photos of a known calibration target. An online camera calibration tool like CalibrX does it in the browser: it detects checkerboard, ChArUco, ArUco or circle-grid patterns, lets you validate each frame, solves pinhole, wide-angle and fisheye models, and shows the error before you export. The result is a set of calibration parameters — focal lengths, optical centre, distortion coefficients, image size, model type and quality metrics — that engineers use to undistort images, measure geometry, align sensors, estimate pose and build 3D reconstructions for robotics, inspection and AR.
Reviewed by the CalibrX engineering team. Last updated September 27, 2026.
How it works
How to calibrate a camera online in 4 steps
- Print a calibration pattern. A ChArUco board is the most forgiving; a checkerboard suits controlled setups. Both, plus ArUco markers and circle grids, can be generated for free. Print at 100% and measure one square.
- Take 15–30 photos. Move the board across the whole image — centre, edges and corners — and tilt it up to about 45°. Keep the focus and zoom fixed: sharp, well-lit photos matter more than many photos.
- Upload and review. CalibrX finds the pattern in every photo, shows the detected corners, and lets you approve or reject each frame before anything is solved.
- Solve and export. Pinhole, wide-angle and fisheye models are solved from the same photos, each with its reprojection error. Download JSON or YAML for OpenCV and ROS, or use the Python SDK.
Capabilities
Everything a camera calibration tool should do
Detection
Auto-detect any pattern.
CalibrX recognises ArUco, ChArUco, chessboard, and circles-grid captures and infers board dimensions, marker sizes, and orientation directly from the images. No CLI configuration, no parameter guessing — drop the captures in and detection adapts to the geometry on screen.
Validation
Validate every detection.
Review per-frame overlays in a fast browser grid. Approve strong detections, reject weak ones, or flag frames with no pattern — without leaving the workflow. The calibration only ever sees the frames you have explicitly validated.
Model selection
Compare every camera model.
Solve the same frames with Pinhole, Pinhole-wide, and Fisheye intrinsics side by side, then pick the model with the lowest RMS error and the cleanest undistortion. CalibrX shows the numbers and the visuals together — no spreadsheet juggling.
Analytics
See where error lives.
Per-frame RMS distributions, residual error maps, and frame-quality trends pinpoint which captures hurt the solve. Use the visual diagnostics to remove outliers and converge on a tighter calibration, run after run.
History
Every run, traceable.
Each calibration is a versioned record — original images, detection settings, validation decisions, and final intrinsics — attached to the camera it belongs to. Roll back, compare versions, or audit the path from raw captures to production parameters.
Exports · SDK
Ship to any pipeline.
Download JSON or YAML calibration files compatible with OpenCV and ROS, or pull intrinsics directly from the CalibrX Python SDK with model-aware undistort helpers. The result drops straight into your existing computer-vision pipeline.
Results
What you get from a calibration
| Output | What it is | Used for |
|---|---|---|
| Camera matrix K | The focal lengths fx, fy and the optical centre cx, cy, in pixels. | Undistortion, pose estimation, 3D reconstruction |
| Distortion coefficients D | How the lens bends straight lines: 5 terms for pinhole (k1, k2, p1, p2, k3), 8 for wide-angle, 4 for fisheye (k1–k4). | Undistorting and rectifying images |
| Reprojection error | The average distance, in pixels, between the detected corners and where the solved model puts them — overall and for every photo. | Judging the calibration and removing bad frames |
| Undistortion preview | One of your own photos corrected with each model, side by side. | Checking the result by eye |
| Calibration files | JSON and YAML (OpenCV FileStorage compatible), plus Python SDK, OpenCV and C++ snippets. | Your code, ROS nodes and pipelines |
Cameras & lenses
Calibrate webcams, machine-vision, action and fisheye cameras
Standard lens
Webcams and phone cameras
Webcam calibration in minutes: standard lenses fit the pinhole model, and a printed A4 board is all you need.
Industrial
Machine-vision and robot cameras
Inspection, pick-and-place and ROS robots: the exported files load straight into OpenCV and ROS pipelines.
Up to ~120°
Wide-angle and action cameras
The wide model adds the rational distortion terms that the basic pinhole model lacks at wide fields of view.
Up to ~180°
Fisheye and dome lenses
Solved with the equidistant fisheye model. Use a ChArUco board: it keeps detecting where a checkerboard curves out of reach.
Camera models
Pinhole, wide, or fisheye — which camera model?
| Model | Typical field of view | Distortion model | Use it for |
|---|---|---|---|
| Pinhole | Up to ~90° | Brown–Conrady (radial + tangential) | Standard rectilinear lenses, webcams, most machine-vision cameras. |
| Pinhole wide | ~120° | Rational (adds k4, k5, k6) | Wide-angle and action cameras where the basic pinhole model breaks down. |
| Fisheye | Up to ~180° | Equidistant (Kannala–Brandt) | Dome and 180° fisheye lenses. |
Not sure which applies to your lens? Read Pinhole, wide, or fisheye? Choosing the right camera model.
Compare
CalibrX vs OpenCV and MATLAB for camera calibration
| CalibrX | OpenCV script | MATLAB Camera Calibrator | |
|---|---|---|---|
| Install | Nothing — runs in the browser | Python or C++ with OpenCV | MATLAB with the Computer Vision Toolbox |
| Writing code | None | Yes, for detection, solving and export | None — a desktop app |
| Fisheye lenses | Yes, next to pinhole and wide-angle | Yes, with the cv2.fisheye module | Yes |
| Reviewing photos | Approve or reject each detection visually | By hand, in your own code | Yes, in the app |
| Output | JSON, YAML and code snippets | Whatever your script writes | MATLAB camera parameter objects |
| Cost | Free to try; licences from $49, one-time | Free and open source | MATLAB and toolbox licences |
Coming from code? See how CalibrX replaces cv2.calibrateCamera scripts and the MATLAB toolbox.
Pricing
Is there a free camera calibration tool?
Yes: the whole CalibrX workflow is free to try. A free account gets 50 credits — about two full calibrations — with detection, frame review, the reprojection error and the undistortion preview. A one-time licence, from $49 with nothing to renew, unlocks the calibration parameters and export files for every camera model. See pricing.
FAQ
Camera calibration software — common questions
Can I calibrate a webcam online?
Yes. Print a checkerboard or ChArUco board, take 15–30 photos of it with the webcam from different angles and upload them. Most webcams fit the pinhole model; CalibrX solves it and shows the reprojection error so you can judge the result.
Do I need OpenCV, Python or MATLAB?
No. Detection, calibration and undistortion all run in CalibrX. You only need OpenCV or Python later, to use the exported parameters in your own code, and the export includes ready-made snippets for that.
Is there a camera calibration app to install?
No installation needed: CalibrX is a camera calibration app that runs in the browser on Windows, macOS and Linux. Take the photos with the camera you are calibrating, then upload them from your computer.
Which calibration pattern should I print?
A ChArUco board for most cameras, and especially for wide-angle and fisheye lenses; a checkerboard for simple, controlled setups. Both can be generated and downloaded for free as print-accurate PDFs.
How many photos do I need?
Between 15 and 30 sharp photos that together cover the whole image, including the edges and corners, with the board tilted in different directions.
What is a good reprojection error?
Lower is better: well under one pixel usually indicates a good calibration. CalibrX also shows it for every photo, so you can remove the frames that pull the result off.
Which files can I export?
JSON and YAML files with the camera matrix, distortion coefficients and image size — the YAML is OpenCV FileStorage compatible — plus Python SDK, OpenCV and C++ snippets.
Keep reading: the complete camera calibration workflow, the documentation and the calibration glossary.