I am using a camera matrix and undistortion coefficients that I found here for some GoPro footage (because I unfortunatly no longer have access to a GoPro to calibrate the footage myself), and it seems to work well enough for 1280 * 960 (4:3), however, I am running into some issues with larger resolutions (3840 * 2880 (4:3)).
According to the openCV documentation:
While the distortion coefficients are the same regardless of the camera resolutions used, these should be scaled along with the current resolution from the calibrated resolution.
Is these referring to the camera matrix? What is the exact procedure to scale the camera matrix? I tried linear interpolation (scalar multiplication of row X with width ratio and row Y with height ratio), the image is better but still doesn't look quite right. Could this be because the video aspect ratio of the calibration footage (16:9) isn't quite the same as my footage? If so, why would the error not scale linearly as well?
E.g.
f_x_target = f_x_calibration * targetWidth / calibrationWidth
c_x_target = c_x_calibration * targetWidth / calibrationWidth
f_y_target = f_y_calibration * targetHeight / calibrationHeight
c_y_target = c_y_calibration * targetHeight / calibrationHeight