The cv2.remap function in OpenCV has inherent dimension limitations, particularly in handling images larger than 32,767 pixels in any dimension due to the use of signed 16-bit integers for internal indexing. This limitation can cause issues when processing large images or performing remapping operations on datasets that exceed these limits.

We are looking for a solution to bypass this limitation by replacing cv2.remap with a more flexible alternative, such as scipy.ndimage.map_coordinates, which does not have the same constraints and allows for full-precision floating-point indexing.

This change would allow for remapping larger images without running into dimension size restrictions.