To find the original unblended frames my python script uses a high-pass filter to look for edges in the frame. Basically, the fewer edges it finds, the more likely it is to be an unblended frame.