chengkunli96/EndoscopeCalibration

Tutorial about how to calibrate stereo camera, and also provide some scripts to simplify this processing.

<style> img[src*="#width-full"] { width: 100%; margin: auto; display: block; } </style>

Calibration for Endoscope

This reporsitory aims to provide some scripts for easily doing calibration for stereo camera. Here, I use endoscope calibration case as example. And the nessary steps could be found below.

Checkerboard Type

Our lab's checkerboard could be found here.

Prepare for calibration

cd scripts

For stereo video file, we do following.

# this script will output the information about this stereo video
python checkStereoVideo.py --file stereo_video_file

# follow the information obtained by the previous script
# do the following, and it will seperate the stereo file into left image sequence and right image sequence
python convertStereoVideo2Images.py --file stereo_video_file  --save img_sequence_dir --leftroi roix roiy roiw roih --rightroi roix roiy roiw roih [--resize resw resh]

For left video and right video (left view and right view are stored in two videos respectively), we do following.

# check the black boundary of the video
python checkMonoVideo.py --file video0_file
python checkMonoVideo.py --file video1_file

# removing the black boundary and save frames as image sequence
pyhton convertStereoVideo2Images.py --file stereo_video_file  --save img_sequence_dir --roi roix roiy roiw roih --rightroi roix roiy roiw roih [--resize resw resh]

There are some other useful script for us to use, like resize an image sequece or select suitable images to calibrate.

# run this for easily select the corresponding images which have the same file name with images in the seleted dir.
python selectCorrespondingStereoImages.py  --ref selected_dir --fromdir source_sequence_dir --todir target_selected_dir

# For every image in the sequece, do resize operation and save them.
python resizeImageSequence.py --dir img_seq_dir --save proc_sv_dir --resize resw resh

Calibrate by Matlab

1. Open matlab find APP/Stereo Camera Calibrator.
2. Add Images, input left and right images sequences dir path. Input your checkerboard size.
3. Select 3 Coefficients, Tangential Distortion, Skew (this option is not necessary). And print Calibrate button.
4. You will see the error result in Reprojection Errors window. Drop the red line in this window to select the outlier which you want to delete.
5. Then right click on the image pair you selected, click Remove and Recalibrate. Please notice, sometimes, error happens and it maybe do not do recalibrate automatically. In this strange case, you can print Calibrate button manually (by check or non-check skew, this button will reactivate).
6. Finally, print Export Camera Parameters/Export to workspace, and make sure the stereo parameters variable is stereoParams and press OK.
7. In the workspace of matlab, make sure the workspace path is in ./scripts/matlab2OpencvYaml, run script save2yaml. And you will find an endoscope_calibration.yaml file in ./scripts/matlab2OpencvYaml/out.

Other instruction

1. The camera parameters will change when you do crop or resize operation: For cropping, cx, cy which change, and cx, cy, fx, fy will change basing on resize ratio in the case of resize. However, baseline parameter will never be changed.
2. The unit of the camera parameters: by our save2yaml.m script, the unit of cx, cy, f is pixel. Differently, the unit of baseline of stereo camera is millimeter.