The code:

• Download  Makefile,  the class file: calibCam.h,   calibCam.cpp,  the drive calibration.cpp.
• Download the image (data) files there.
• Type:  make.
• The member function directCalibration()  implements the direct calibration.
• The member function accurateE() gives the accuracy test.
• The member function projectionE()  gives the projection test.

Usage:

• (I compiled and ran my program in RedHat8.0 and OpenCV0.9.4).
• After compiling, type: ./calibration <file-number> .....
• (for example, if using 01, 03, 05, 07, 09 image/data-file, type:  ./calibration 1 3 5 7 9), and following will be occurred:

[b_yi@chicken prg3]$ ./calibration 1 3 5 7 9

 input file number for accuracy test ?  5
accuracy error:  97.1017

 input file number for projection test ?  10
projection error :  125.508

• And also an image will be displayed and saved showing the original tested image (given by the requested input) with green cross at the estimated / given pixel location and red cross at the computed pixel location (with the computed calibration parameters). (For some locations which match so well that  you can only see either the red crosses or the green ones with the other occluded.)
  

Results:

1. When using images 01, 03, 05, 07, 09:

• Accuracy test error (accumulated error for all the 96 points in one image) for image 05 is : 97.1017.  
• The parameters:

Extrinsic Parameters: rotation matrix:
0.685203  0.033461  -0.727583
-0.108544  -0.983098  -0.147433
-0.720219  0.179996  -0.66999
Extrinsic Parameters: translation vector:
-164.661
94.392
724.197
Intrinsic Parameters:
731.056  0  325.737
0  734.467  246.272
0  0  1

• And the effect can be seen in following image:
  

• Projection test error (accumulated also) for  image 10 is:  125.508.  
• The parameters (extrinsic only, intrinsic same as above):

Extrinsic Parameters: rotation matrix:
0.999711  0.000843717  -0.0240213
-0.0113629  -0.864067  -0.503249
-0.0211806  0.503376  -0.863808
Extrinsic Parameters: translation vector:
-159.31
78.8653
576.958

• And the effect can be seen in following image:

   
2. When using images 01, 02, 03, 04, 05, 06, 07, 08, 09:

• Accuracy test error (accumulated error for all the 96 points in one image) for image 07 is : 122.602.  
• The parameters:

Extrinsic Parameters: rotation matrix:
0.920533  -0.0161807  0.390329
0.0540686  -0.984249  -0.168314
0.386905  0.176043  -0.905159
Extrinsic Parameters: translation vector:
-142.245
69.0835
553.8
Intrinsic Parameters:
732.631  0  324.235
0  735.863  245.581
0  0  1

• And the effect can be seen in following image:
  

• Projection test error (accumulated also) for image image 15 is:  151.227.  
• The parameters:

Extrinsic Parameters: rotation matrix:
0.981816  -0.162489  0.0981578
-0.186909  -0.917877  0.350096
0.03321  -0.362076  -0.931557
Extrinsic Parameters: translation vector:
-132.026
77.2353
702.764

• And the effect can be seen in following image: