This repository contains a code that can be used to calibrate an ESP32 camera. The implementation of the cam package that is used here to get frames from the camera may or may not need to be modified to comply to the camera you are using.
After cloning the repository simply run pip install -r requirements.txt and you're all set to go.
Run script calibration.py - it will take all the example images from screenshots/ directory and perform a camera calibration based on them. After the calibration the program will draw a raw realtime stream of frames from the camera and below it a rectified version of the same stream after applying the calibration.
First off you need to print the chessboard calibration pattern. Then run the image_capture.py script and using spacebar capture around 20 images of the chessboard from different angles and using different parts of the camera screen. These images will be saved to screenshots/ directory which already contains a few sample images of the chessboard, so if you want to run a calibration only based on your new images you have to delete them prior to this. Secondly run calibration.py. All the images used for the calibration will be displayed one by one with the chessboard pattern drawn on them. After this process you will get stream of frames from your calibrated camera side-by-side to the image from before the calibration for comparison.
Below are presented images used for calibration:
After running calibration.py on those images we get side-by-side view of images before and after calibration (difference is barely visible because the camera used here had already pretty good image prior to calibration):
