molerus
21st February 2003, 23:55
Hi!
For 2 days I'm dealing with problematic scene taken from Terminator. I combined PixieDust with Convolution3d and it appeared, that even with the quality set to 100% I get blocks. Previously I was using h.263 quant matrix, but Didee advised me to use MPEG quant (thx). In fact it solved the problem (well, nealy solved :rolleyes: ).
But for the rest of the film I would prefer using the h.263 quant matrix since it's more pleasing to my eyes, and less bits consuming (not always true, I found that for high motion scenes MPEG quant matrix may give smaller frames :confused: ).
I ported NanDub's code asessing LumaNoise into a VDub filter. I'm interested in using MPEG matrix ONLY for low luma noise scenes.
My question is if it's possible to force XviD to use certain quant matrix (i.e. scenes with low luma noise and/or high motion --> MPEG, otherwise --> h.263)??? Is this possible, either through stats file or differently??
I have read that Modulated matrix behaves in such way, that it switches between h.263 matrix for quant 2-3, and MPEG for quant above 3. But frankly I can't believe it, since when I did a two pass encoding on mentioned scene, setting final size way above the first pass size, so the quant was always 2, XviD seemed to use MPEG matrix either way (the final size was nearly the same as 100% using the MPEG matrix). Maybe the decision process works other way??
BTW - what are the kBlk, mBlk and uBlk in stats file???
For 2 days I'm dealing with problematic scene taken from Terminator. I combined PixieDust with Convolution3d and it appeared, that even with the quality set to 100% I get blocks. Previously I was using h.263 quant matrix, but Didee advised me to use MPEG quant (thx). In fact it solved the problem (well, nealy solved :rolleyes: ).
But for the rest of the film I would prefer using the h.263 quant matrix since it's more pleasing to my eyes, and less bits consuming (not always true, I found that for high motion scenes MPEG quant matrix may give smaller frames :confused: ).
I ported NanDub's code asessing LumaNoise into a VDub filter. I'm interested in using MPEG matrix ONLY for low luma noise scenes.
My question is if it's possible to force XviD to use certain quant matrix (i.e. scenes with low luma noise and/or high motion --> MPEG, otherwise --> h.263)??? Is this possible, either through stats file or differently??
I have read that Modulated matrix behaves in such way, that it switches between h.263 matrix for quant 2-3, and MPEG for quant above 3. But frankly I can't believe it, since when I did a two pass encoding on mentioned scene, setting final size way above the first pass size, so the quant was always 2, XviD seemed to use MPEG matrix either way (the final size was nearly the same as 100% using the MPEG matrix). Maybe the decision process works other way??
BTW - what are the kBlk, mBlk and uBlk in stats file???