Last modification which served me to optimize the input
information for the actual detection was the reduction of
number of pixels in the resulting image. By this we once
again significantly reduced the burden on the system and we
got rid of another error distortion.