Zhang's camera calibration

Implementation of Camera calibration method as presented by Zhengyou Zhang from Microsoft in his paper, A flexible new technique for camera calibration

Results

Mean reprojection error(pixels) for calibration images before & after optimization

Image	#1	#2	#3	#4	#5	#6	#7	#8	#9	#10	#11	#12	#13
Before	1.36	0.84	1.38	18.22	1.09	4.55	2.90	4.19	5.30	4.86	22.52	12.02	41.11
After	0.35	0.49	0.84	1.14	0.18	0.44	0.08	0.21	0.34	0.29	0.62	0.86	0.88

Intrinsic matrix and distortion coefficients' estimates:

1. OpenCV reference:

   camera matrix:
      [[2.042729e+03 0.000000e+00 7.643600e+02]
      [0.000000e+00 2.035016e+03 1.359026e+03]
      [0.000000e+00 0.000000e+00 1.000000e+00]]
   
   distortion coefficients(k1 and k2):
      [0.290493410 -2.42737867]
   

2. Initial estimates:

   camera matrix:
      [[2.061892e+03 −2.850592e + 00 7.760024e + 02]
      [0.000000e + 00 2.047799e + 03 1.363240e + 03]
      [0.000000e + 00 0.000000e + 00 1.000000e + 00]]
   
   distortion coefficients:
      [0.0, 0.0]
   

3. Optimization without considering distortion coefficients:

   camera matrix:
      [[2.058478e+03 −1.454986e + 00 7.540409e + 02]
      0.000000e + 00 2.049441e + 03 1.354108e + 03]
      0.000000e + 00 0.000000e + 00 1.000000e + 00]]
   
   distortion coefficients:
      []
   

4. Optimization with distortion coefficients:

   camera matrix:
      [[2.048509e+03 −1.830419e + 00 7.587442e + 02
      0.000000e + 00 2.040725e + 03 1.345145e + 03
      0.000000e + 00 0.000000e + 00 1.000000e + 00]]
   
   distortion coefficients:
      [0.17310107 -0.75331797]
   

Steps to run the code

1. Install dependencies

   python -m venv venv
   source venv/bin/activate  # ./venv/Scripts/activate for windows
   pip install -r requirements.txt
   

2. Run script

   python Wrapper.py