I am working on a project involving real-time comparison of two images taken from different angles. The goal is to accurately detect and align the angles between these images to facilitate further image processing tasks. I am using Python libraries including Pillow, OpenCV, and SSIM for this purpose.

I used the opencv.TM_CCOEFF_NORMED function for matching the invoice template. However, the template matching functionality does not yield accurate results. the matching accuracy is quite poor, at only 50%. Could you please advise on what I should include or change in the process to increase the accuracy?