+ Reply to Thread
Results 31 to 34 of 34
As I think I mentioned somewhere in the thread where I posted that script, photographer's flashes sometimes include several "pre-flashes." If there is at least one non-flash frame on either side, the script will repair those as well, but if they are fast enough that they happen over consecutive frames, the script will fail. Put another way, the bad frame must have a good frame on either side for the script to detect that it doesn't belong and therefore needs to be removed.
I am pretty happy with that script. You can sometimes improve its effectiveness if you take a few minutes to look at the metrics and then change the detection threshold. For most videos, you can actually err on the side of having it find too many "bad" frames. While this results in some good frames being replaced, the motion estimation is usually so good that you don't notice.
Now if I could just finally figure out how to "perfect" the script I wrote which fixes bad streaming and bad captures where you end up with dropped frames followed by dups. It actually works as well as this one, but my method requires breaking down the video into groups of frames. If the dup/drop pair happens on either side of the boundary between groups, it gets missed. It removes 80-90% of the problem, but if I can ever figure out a way around the problem, it should get 99+%.
Need help "perfecting" script to delete drops and dups