Quantcast KVCD Template Comparison? - digitalFAQ.com Forums [Archives]
  #1  
06-19-2002, 11:26 PM
Daagar Daagar is offline
Free Member
 
Join Date: May 2002
Posts: 158
Thanks: 0
Thanked 0 Times in 0 Posts
Dunno if this info will be useful to anyone, but since I bothered to do the test, I figured I'd share the results.

Parameters:
Source: 640x272 DivX file (hush, I know)
Codec: ffdshow <-- An _excellent_ Divx/Xvid codec. Use it!
Full postprocessing, 100% for all tests [fullprocessing]
Resizing done with ffdshow (bilinear for downsize, bicubic for upsize)
Encoder: TMPGEnc 2.54 (Normal motion search on all tests)
Templates: Kwag 352x480 and Kwag 704x480

Variables:
ffdshow Sharpen (default settings) [sharpen]
TMPGEnc Noise Reduction (default settings) [denoise]
Both/Neither of the above.

I encoded 32 secs (790 frames) each time, changing the parameters above. Size measurement are what 'Windows Explorer' reports when looking at the file size. This particular clip comes from Lord of the Rings, the scene where Gimli tries to break the One Ring with his axe. There is a mix of high/low motion, panning, and rapid scene changes in this 30-sec, so hopefully it is somewhat representative of 'real' results.

Results:
352x480 [fullprocess] 4,133kb (~3-4hours/90min)
352x480 [fullprocess] [denoise] 3,718kb (~6-8hours/90min)
352x480 [fullprocess] [sharpen] 3,947kb (~3-4hours/90min)
352x480 [fullprocess] [denoise] [sharpen] 3,777kb (~6-8hours/90min)
704x480 [fullprocess] 3,427kb (~5-6hours/90min)
704x480 [fullprocess] [denoise] 3,312kb (~15hours/90min)
704x480 [fullprocess] [sharpen] 3,480kb (~6-7hours/90min)
704x480 [fullprocess] [denoise] [sharpen] 3,364kb (~15hours/90min)

Interesting side note: The .avi file had a VBR audio track. I left the soundtrack in while doing these tests (doh!). The 704x480 template would have terrible a/v sync, but oddly.. the 352x480 stayed in perfect sync! Not sure how useful this is, but was strange to see. I normally pull the audio out with ssrc/tooLame and remux.

Summary:
The 'ffdshow' sharpen filter claims that it tends to reduce noise, and as seen above the filesizes _are_ slightly lower when using the sharpen filter. Not as much as using tmpgenc noise reduction, of course. However, the sharpen filter _adds no additional time_ to the encoding process, whereas noise reduction adds LOTS of time.

Surprisingly (to me), the 704x480 template came out smaller! CQ of 50 vs. 70 makes quite a difference I guess. Playback of these files is quite subjective. It was hard to tell a difference between any of them, even the ones of different resolution. On my TV (25"), if I had to choose one, I'd say the 704x480 [fullprocessing] [sharpen] looked the best.

One thing that was noticible between the two was the color. The 352x480 samples seemed too 'yellow' in comparison with the 704x480 templates. I'm sure this could be corrected via color filters, but heck if I know how to tweak them correctly. Obviously, 352x480 was slightly blurrier than 704x480, but on my TV it wasn't a HUGE difference.

Another shock was that 704x480 _did_ look better, considering I had to upsize my source! Granted, it was only 640->704, but I expected worse quality going up. Instead, it was on-par with going down.

Doing noise reduction, while it did shrink the file size a bit, adds so much additional processing time that I'm unsure it is worth it. I could see no visual difference (remember my source!), and since I think I prefer 704x480, 6hours per 90min vs. 15+hours/90min.... wow. And this was just with normal motion search. Setting it to 'high' would incur even stiffer penalties.

I really really like ffdshow as a DivX codec. It plays all DivX/Xvid/MPEG4 formats, and allows all the various crop/aspect/resize utilities that I normally use AVISynth for. The only thing I think it is missing is an equivalent of TemporalSmoother. I guess a next test should be a few tests between encoding with ffdshow and AVISynth with various filters between the two.

If anyone had any other ideas of what I should do to make a 'fair' comparison, let me know. Again, I have no idea if any of the above is useful to anyone, but I figured I'd share. It is probably info everyone already knows
Reply With Quote
Someday, 12:01 PM
admin's Avatar
Site Staff / Ad Manager
 
Join Date: Dec 2002
Posts: 42
Thanks: ∞
Thanked 42 Times in 42 Posts
Reply




Similar Threads
Thread Thread Starter Forum Replies Last Post
Windows XP SP2 vs SP1 Performance Comparison jorel Computers 11 10-20-2004 04:08 PM
Avisynth: Speed comparison shibumi Avisynth Scripting 7 05-10-2004 08:14 AM
DVD-Rom Installation and comparison help bigggt Blank Media / Burning Discs 7 01-25-2004 09:34 AM
KVCD Template Comparison: Mpeg-1 vs. Mpeg-2 ? Jellygoose Video Encoding and Conversion 3 12-12-2002 04:41 AM
KVCD: Comparison samples Daagar Video Encoding and Conversion 3 06-14-2002 07:52 AM

Thread Tools



 
All times are GMT -5. The time now is 07:29 PM  —  vBulletin © Jelsoft Enterprises Ltd