I'm working with the archaic observational data that are greyscale images like the one attached. The total set consists of ~10000 images taken a long time ago. Each image contains a N-E coordinate axes usually drawn in the corners of the image. The goal of the project is to rotate images by a certain angle in such way that N-axis mathes the y-axis and the E-axis to match the negative x-axis. For example, this image needs to be rotated by ~45 degrees in a clockwise direction to achieve this goal.

Does anyone have any suggestion in how could I detect the coordinate lines, perform some fitting/calculating procedure to find the rotation angle and rotate the image. Rotation should be straightforward with known angle, no need to comment on that xD.

Long story short, the aim is to find the rotation angle based on the arrows position and rotate the image, and then repeat the procedure for the whole dataset in Python 3.7+

All comments and suggestions are welcome.