Pal retail dvd with blends, looks bad

[poisondeathray]

I haven't look at your attachment yet - but was it the result of MT or the script in general ? Was this part of the full encode, or did you only test out a small section

The problem with srestore (and some temporal filters) is if you don't start at the beginning and go to the end, the results will be different if you start from a seeking to a point somewhere in the middle - it's a pain to debug. Each time you start in a different place , you might get different results. Same with MT.

You would have to examine a large set of frames around that site (preferrably beginning to end, but sometimes a large subset is good enough), try tweaking the script with/without MT to see if that's the cause (each time you need to encode linearly from the beginning - because of the non linear seek problem - you can't just preview a script in a middle section, you have to go back a long distance because those filters might choose different frames each time)

Also note, these filters aren't perfect, they might make wrong decisions in picking frames, even if there was a potential better match. There might be some settings you can tweak in srestore to get better decisions

[jagabo]

Decimation filters often have problems when there are only small motions between frames.