Calibration Process
 |
|
Calibration Process
Stabilizer Calibration Tree
This page organizes the stabilizer calibration workflow into a step-by-step tree. Each node links to a detailed section below so you can jump quickly between setup, calibration, synchronization, and tuning tasks.
The calibration of the camera is crucial to map the rotations into the pixel-space reference system. The IMU calibration is needed to determine the proper orientation of the IMU with respect to the camera, and how laggy the camera is. And the tuning allows quantitatively evaluate the quality of the stabilization using jitter and SSIM metrics.
The RidgeRun Video Stabilization library proposes a series of tools for remote calibration using a web tool or local calibration in case of a custom board is needed. The camera calibration can be done using the web tool, a Docker container and an on-target tool. The IMU calibration can be done using the web tool and an on-target tool. The tuning tool can be used through the web tool and a local script.
Suggested flow: pattern preparation → camera calibration → IMU orientation calibration → timestamp offset calibration → stabilizer tuning.
-
Stabilizer Calibration Workflow
-
Step 1: Get a 6x9 calibration pattern
25 mm squares -
Step 2: Camera calibration
Intrinsics + distortion parameters- Capture multiple views of the pattern
- Estimate focal length, principal point, distortion
- Validate reprojection error
-
Step 3: IMU calibration
Camera ↔ IMU rotational alignment- Record live motion data
- Estimate sensor-to-camera rotation
- Estimate time delay between streams
-
Step 4: Tune stabilizer properties
Filter strength, crop, smoothness, latency- Test on live
- Balance stability vs motion responsiveness
- Final validation
-
Step 1: Get a 6x9 calibration pattern
Step 1: Get a calibration pattern
Prepare a checkerboard calibration target with 6x9 inner corners and 25 mm square size. Print it on a flat, dimensionally stable surface, or mount it on a rigid board to avoid bending.
- Verify the printed square size with a ruler or caliper.
- Ensure the board remains flat during data capture.
- Use good lighting so the corners are detected reliably.
Step 2: Calibrate the camera
Use the calibration pattern to estimate the camera intrinsic matrix and distortion coefficients. Capture the pattern at different positions, scales, and orientations across the image.
- Capture many frames with the board covering different image regions.
- Include tilted views to constrain focal length and distortion better.
- Compute intrinsics, distortion, and reprojection error.
- Reject blurry or poorly detected images.
Typical outputs: the camera intrinsic matrix (3x3) and distortion parameters (4 or 5, depending on the lens model).
Back to top - Camera Calibration ProcedureStep 3: Calibrate the IMU
Estimate the rotational alignment between the IMU frame and the camera frame. This is needed so that gyroscope or fused orientation data can be interpreted correctly by the stabilizer. Also, estimate the time offset between the camera frames and IMU samples. Even with a good spatial alignment, poor temporal alignment can cause lagging or incorrect stabilization behavior.
- Make sure to attach rigidly the IMU as centred to the visual axis as possible.
- Compare visual motion estimates with IMU angular motion.
- Find the time offset or camera lag.
Output: orientation triplet and timestamp lag.
Back to top - IMU CalibrationStep 4: Tune stabilizer properties
After geometry and synchronisation are calibrated, tune the stabiliser behaviour itself. This step adjusts the trade-off between smoothness, responsiveness, crop, and visual artefacts.
- Adjust smoothing coefficient and technique.
- Set crop margin or field-of-view protection.
- Fine-tune the offset and orientation.
- Validate on real clips with walking, panning, and high-frequency vibration.
Final goal: stable output without excessive latency, wobble, or overly aggressive cropping.
Back to top - Tuning