From OpenCV to StereoComplex

You already know cv2.calibrateCamera and cv2.stereoCalibrate? Here is the equivalent path in StereoComplex, plus what you gain.

3-minute quickstart

from pathlib import Path
import stereocomplex as sc

# 1. Define your ChArUco board (same as OpenCV)
board = sc.CharucoBoardSpec(
    squares_x=11, squares_y=7,
    square_size_mm=39.07, marker_size_mm=27.35,
    aruco_dictionary="DICT_4X4_1000",
)

# 2. Compare OpenCV raw vs Ray2D-refined in one call
report = sc.compare_opencv_stereo_calibration(
    left_dir=Path("my_data/left"),
    right_dir=Path("my_data/right"),
    board=board,
)
print(f"Raw stereo RMS:    {report['raw']['stereo_rms_px']:.3f} px")
print(f"Refined stereo RMS: {report['refined']['stereo_rms_px']:.3f} px")
print(f"Improvement:        {report['improvement_px']:.3f} px")

# 3. Export to OpenCV format
K1, d1, K2, d2, R, T = report["refined_result"].to_opencv()

Translation table

OpenCV

StereoComplex

cv2.aruco.CharucoBoard

sc.CharucoBoardSpec(...)

cv2.aruco.detectMarkers

sc.detect_charuco_corners(image, board)

cv2.calibrateCamera

sc.calibrate_opencv(left_dir, right_dir, board)

Raw corners

method2d="raw"

Refined corners (Ray2D)

method2d="rayfield_tps_robust"

K1, d1, K2, d2, R, T

result.to_opencv()

Compare raw vs refined

sc.compare_opencv_stereo_calibration(...)

What you gain

Feature

OpenCV alone

With StereoComplex

Corner accuracy

Raw ChArUco

Ray2D-refined (~2× better)

Non-central optics

Not supported

Full Zernike rayfield pipeline

Model identification

Manual

Automatic BIC-based selection

Optics catalogue

None

6 families (pinhole→CMO→Greenough→exotic)

Next steps

  • I want better corners before OpenCV → sc.compare_opencv_stereo_calibration()

  • I want a central ray-based model → sc.calibrate_central(left_dir, right_dir, board)

  • I suspect non-central optics → sc.calibrate_noncentral(left_dir, right_dir, board)

  • I want to identify my microscope optics → sc.identify_optics(target_field, candidate_specs, K, image_size)

See also