Shandra
3rd July 2003, 04:20
Hi Folks,
I am really not sure where to put this (crazy/stupid) idea... So mods... feel free to move it to the place you feel apropriate for it!
But (inspired by some lectures I had on GeoStatistics recently) lately with my DVD backups I had an idea that I do not know if that is what multiple pass actually means or if it would be a different concept...
consider ONE codec with known settings.... consider a filter which predicts the encoded frame (with maybe artifacts) and compares it to the actual source frame... then the filter modifies the source frame with derived parameters from the predicted encoded frame in such a way that the resulting artifacts or sources for artifacts got blended in the actual encoding in such a way that they are less visible in the result....
Would such a thing be possible? I know this is a step away from nearly lossless but somehow I do believe that you can meddle with the source to let it become a codec-friendly in such a way that visible quality can be preserved during encoding (at least to more % than it is now)... or is that what 2pass and beyond already does?
I am really not sure where to put this (crazy/stupid) idea... So mods... feel free to move it to the place you feel apropriate for it!
But (inspired by some lectures I had on GeoStatistics recently) lately with my DVD backups I had an idea that I do not know if that is what multiple pass actually means or if it would be a different concept...
consider ONE codec with known settings.... consider a filter which predicts the encoded frame (with maybe artifacts) and compares it to the actual source frame... then the filter modifies the source frame with derived parameters from the predicted encoded frame in such a way that the resulting artifacts or sources for artifacts got blended in the actual encoding in such a way that they are less visible in the result....
Would such a thing be possible? I know this is a step away from nearly lossless but somehow I do believe that you can meddle with the source to let it become a codec-friendly in such a way that visible quality can be preserved during encoding (at least to more % than it is now)... or is that what 2pass and beyond already does?