PDA

View Full Version : xvid nice, but


sumpm1
23rd February 2002, 16:14
I think that there is still further development on the open divx, but I am a 1 pass type of person that goes for 1 cd rips, and I have been using divx4 and has worked for me. I have seen the comparisons of the codecs here, and that is if you freeze frame. When you are watching the movie, I cant tell the difference usually between any of the divxs, but it becomes more apparent when the bitrate gets much lower. I also saw that divx4 is better at changing the bitrate, as to not get blockiness in fast scenes or high detail scenes.And divx4 gives better compression than div3.

Also, Divx4 claimed to have nothing to do with divx3 or mpeg4, is this true?

serbersan
23rd February 2002, 16:43
The main problem of divx4 besides the undersized files it's that the codec blurs the image too much. This is why you don't see macroblocks but this is why divx3 and xvid have more crispness or sharpness.

and about divx3 if you know all it secrets you could do better rips than with divx4 at least with 1CD rips. But divx4 offers really good results without any tweaking.

About seeing the movie you can't tell the difference between codecs it's a partial truth at least for me. I notice more sharpness/crispness and better overall image (but in freeze frames is not as visible like watching the movie) in my xvid rips than in divx4 and xvid it's as easy to use like divx4.

I believe the dead of divx3 is near,near.

NeVeRLiFt
24th February 2002, 02:54
I look forward to XVID. But I still use Nandub.
I guess we will just have to wait and see what this new M$ codec is gonna be like and see if its hacked or usable by us.... also wanna wait and see what happens with these DVD units that are gonna use Mpeg4 ;) wanna see what they use and hopefully start using it :D.
The future is bright for mpeg4 no matter how it turns out. There is still DivX5.0 that is supposed come out to.

/me sits back and waits

dragoman
24th February 2002, 09:17
Hi,

I still use DivX 3.11 with Nandub. Quality is better than DivX 4.xx and I have less problems. Divx 4.xx seems still buggy to me.

As for Xvid, I wasn't really impressed when I compared with Nandub-encoded movies.

We'll see if DivX 5.0 is any better...

dragoman

kastro68
24th February 2002, 12:06
Originally posted by serbersan
The main problem of divx4 besides the undersized files it's that the codec blurs the image too much. This is why you don't see macroblocks but this is why divx3 and xvid have more crispness or sharpness.





True...but from what i've heard, some ppl use a lower resolution while encoding because it acts as a sort of noise filter. By lowering the resolution it also results in loss of detail, slightly blurring the image.

I find that with divx 4.12, since it already blurs...you should try encoding using a higher resolution than you would otherwise than if you were working with nandub.

I find 4.12 still useful for movies that are really really hard to compress to one cd.


But Xvid it seems works really really well when you use 2cds, and might work better than 4.12 if you use H.263 for 1cd encodes.

-h
24th February 2002, 13:03
DivX4 uses H.263 quantization, but removes small DCT values (i.e. doesn't quantize them, but removes them completely) to lower bitstream size. This was called 'thresholding' in XviD but has been disabled. It has the effect of blurring the picture, as you lose small coefficients which are usually high-frequency, noise-like values.

Despite this additional filtering from DivX, XviD is still getting a smaller file size with identical quantizers. Anyway, if you compressed the same video with DivX and XviD, using identical quantizers for each frame, XviD will result in a smaller file size and a more detailed picture, despite the fact that H.263 was used by both codecs. I suppose you could emulate this in XviD by applying a smoothing filter to the movie, but I'm not sure why you'd want to..

Try it and see - do a constant quantizer encode in XivD and DivX, with the same quantizer. What follows is a fundamental test of how well the codecs determine matches between the current and previous frame (among other things).

-h