Avisynth: Very low CQ for PAL video
The movie is 104 min long and contains very few "action" scenes. Still I can't get the CQ past 56! I usually encode action movies about this length with a CQ around 70.
There is something funny about the DVD though. The picture looks almost like a bad VCD (note: VCD NOT KVCD :wink: ) when you look closely. Some blocks in the background and stuff like that. Is there anything I can do about it? Filter suggestions please. I don't really feel like putting this (short) movie on two CD-R's. |
You could try turning on the postprocessing settings in MPEG2DEC3.
MPEG2Source("path\video.d2v",cpu=4) will remove at least some of the blocks. If the source is interlaced, you'll have to use MPEG2Source("path\video.d2v",cpu=4,ipp=true). |
try this with your own values and whatever you want to do.
AviSource("C:\unzipped\movies/resident evil.avi", false) BlindPP(CPU=4) LanczosResize(352, 334, 4, 0, 568, 304) AddBorders(8, 73, 8, 73) Undot() deen("a3d",1,12,14) ATC(2,3,5,0.5,false) TemporalSoften(2,7,7,3,2) MergeChroma(blur(1.5)) Asharp(1,4) DCTFilter(1,1,1,1,1,1,0.5,0) Resident Evil is 93mins. long and i got a CQ of 77 with that. so you can just try it out if you want, but if you don't want to use that, just use BlindPP then, because BlindPP gets rid of alot of DCT Blocks. :D |
No need to use BlindPP for DVD sources. The built-in postprocessing in MPEG2Source does a better job with them.
|
Thanks guys!
I will start by trying Boulder's suggestion then I will try muhali3's suggestion. BTW is BlindPP a plugin or an AviSynth function? Do I have to specify the BlindPP parameters if I have an interlaced source? |
Quote:
The author of mpeg2dec3 simply wanted to provided a mean to make pp even on non mpeg2 source. But do not expect him to code two different procedures for that ! |
Sorry Boulder. I wrote my reply before I got to read your last post...
OK, I will go with your suggestion Boulder. The other cpu settings, can they produce even better results or do your recommend always using cpu=4 DVD sources? |
@Dialhot.
Hi there. Testing your script right now :wink: Looking good so far :D |
Quote:
(from the MPEG2DEC3.dll docs) Code:
BlindPP(clip, int "quant", int "cpu", str "cpu2", bool "iPP", int "moderate_h", int "moderate_v") |
BTW how does this "deblocking" effect the image quality? Except for getting rid of some of the blocks of course :?
|
Okay, the testing is now complete (I used prediction in CQMatic) and here are the results:
Script WITHOUT 'cpu=4, iPP=true' : CQ56,27 Script WITH 'cpu=4, iPP=true' : CQ56,26?! The "clean" script without "cpu=4, iPP=true" got the highest CQ!? Sounds strange - but I double checked and it's true! The script with "cpu & iPP" has a cleaner look though. But still... Is there any filter(s) I can use to clean up the source even more? I know of VagueDenoiser but I think that filter is only for getting rid of noise. Would it be useful in this case? |
vaguedenoiser is extremely good and preserves the detail very well too
|
Hi!
Quote:
You perform a lot of denoisers but with temporalsmoother set to values like (2,7,7 ...) and above, its a littly risky to sharpen afterwards. Cause TempSmoother with higher values (if not already the most is cleaned by using deen and ATC) could cause an effect like a window in front of the picture with static grain on moving scenes. If this happens, this will not be eliminited by chromaBlur .. it will be even also sharpened by using asharp afterwards. So just do a testing on your bad material, if this effect occurs try one step without using Tempsmoother. :wink: |
Quote:
|
so should i also take out the chromablur :?:
|
you should also change the "a3d" in deen to "a2d", because a2d is spatial-only, and that is good because ATC can take care of the temporal cleaning.
|
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.