Details Lost with Optimal Script?

J-Wo · #1 04-28-2004, 08:15 AM

hey guys. i've been trying v4 of optimal script for mpeg-4. I notice that i lose some details in my image. surfaces appear washed out, edges appear softer. I use CCE to encode mpeg-2 for KDVD compatability, with a Q of 15 and bitrate range of 64-4500. I often find that using NO filtering (resize only) produces a better image from my files, at the sacrifice of diskspace.

Anyone got some suggestions? Is there a modification to these scripts to not do such heavy filtering, just a light cleanup of the image to save some space? Or is it just me who is complaining...?

incredible · #2 04-28-2004, 09:36 AM

EVERYTHING depends on the source quality of a mpeg4!!

The optimal script for very well encoded (and higher avg bitrate used) mpeg4 streams filters a bit heavy. So u should use the regular MA as its also optimal for high quality YV12 sources.

J-Wo · #3 04-28-2004, 02:43 PM

okay well i find that dialhots scripts generally make my resulting image too soft compared to no filtering at all. Are there any tweaks I could make to the scripts to not lose so much sharpness to the image?

Boulder · #4 04-29-2004, 05:08 AM

I ended up removing the TemporalSoften line and the DCTFilter line. Pleases my eyes more without than with them in the script.

incredible · #5 04-29-2004, 06:48 AM

DCT Filter is very tricky and there have been some user issue reports.

As I know Phil deleted the DCTFilter lie as at 352 width it makes the video really blurry!

Better is to enter the matrix and to set the last two collums of each DCT matrix (intra and non-intra(or inter)) to 99

xx-xx-xx-xx-xx-xx-99-99
xx-xx-xx-xx-xx-xx-99-99
xx-xx-xx-xx-xx-xx-99-99
xx-xx-xx-xx-xx-xx-99-99
xx-xx-xx-xx-xx-xx-99-99
xx-xx-xx-xx-xx-xx-99-99
xx-xx-xx-xx-xx-xx-99-99
xx-xx-xx-xx-xx-xx-99-99

This also cuts high fequencies but "could" be not that agressive like DCTfilter does .... we should test this.

J-Wo · #6 05-02-2004, 04:09 PM

hey boulder, I gave your suggestion a shot and I really like it! There's hardly a difference in compression (perhaps a little less) but encoding is a lot faster. And the resulting image IS a lot less soft. Not too sure how to implement the suggestion that inredible made though... Any other tips you guys have?