Fix the code for consistent numpy array usage and passing across the different steps in the DIANNA pipeline 

Currently, the code is using several python packages each transfering the image in a different (numpy) format. By using one package (e.g., sk-image or cv2) consistenty throughout all the code we avoid errors and wrong output and ensure that we obtain reliable results