The origin paper is from Jae-Yeong Lee, “Zero-Shot Calibration of Fisheye Cameras,” arXiv, 2020, https://arxiv.org/abs/2011.14607
How to use this code is described below:
usage: zero_shot_calib.py [-h] [-W WIDTH] [-H HEIGHT] [-hf HFOV] [-vf VFOV] [-d DRAW]
find proper fisheye camera focal length with spec
options:
-h, --help show this help message and exit
-W WIDTH, --Width WIDTH
Width information of image frame
-H HEIGHT, --Height HEIGHT
Height information of image frame
-hf HFOV, --Hfov HFOV
Horizontal field of view angle. It should be float
-vf VFOV, --Vfov VFOV
Vertical field of view angle. It should be float
-d DRAW, --draw DRAW draw plot or notexample usage:
python3 zero_shot_calib.py -W 1920 -H 1440 -hf 122.0 -vf 94.0Then it will show you the focal length in the terminal.
After that, if you want to undistort the image, you can use undistort.py
usage: undistort.py [-h] [--input_image INPUT_IMAGE] [--focal FOCAL] [--output_image OUTPUT_IMAGE]
options:
-h, --help show this help message and exit
--input_image INPUT_IMAGE
--focal FOCAL
--output_image OUTPUT_IMAGEexample usage:
python3 undistort.py --input_image 20210812_084000_000_0400.png --output_image undistort.png --focal 1021.92then it will save the undistorted image.
Example result:
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent