Another topic on blocknoise...

[ChrO]

Hi All
I know a lot of posts and guides have described blocknoise, and what to do to avoid it. And i think i have read it all...
But i still sometimes have the problem!

I capture cartoons in picvideo mjpeg codec, 720x576 quality 18.

I encode them in Tmpgenc, loading DVD PAL template, Using CQ 85, min bitrate of 2000 and max 8000, i have enabled "soften blocknoise", i also use the noise filters, the qualty is set to "high".

I then use EVE to add transitions between them (it only encodes the frames with the transitions)

Finally i author with Tmpgenc DVD author, adding menus etc and burn it.

But in the very fast moving scenes there are still these ugly blocks, in the black and white cartoons the blocks can be very viewable, black on white!
In the rest of the movie the quality is very high.

Is it because i have the limit of 8000 max bitrate, or is it sufficient? (I could try 9000, i only use 384 for the mp2 stream)
Or is it tmpgenc that cant properly detect scenechanges?
Is it something i have not thought about?
Really hope someone can help, thanks in advance!
Best regards.
Chr

[bugster]

A couple of things for you to try:

First, add your transitions to the avi BEFORE encoding to mpeg.

Second, althought TmpGenc's CQ mode is very good, 2-pass VBR has got to be worth a try.

[SatStorm]

Yes, 2 Pass VBR is better CQ (1 Pass VBR). But with TMPGenc, 2 Pass don't worth the time to encode, compare CQ mode... Minimum to no difference, at almost double the time....

So, if you can, give 2 Pass VBR a try with other encoders, like the latest mainconcept or CCE.

[W_Eagle]

Yes, 2 Pass VBR is better CQ (1 Pass VBR). But with TMPGenc, 2 Pass don't worth the time to encode, compare CQ mode... Minimum to no difference, at almost double the time....

So, if you can, give 2 Pass VBR a try with other encoders, like the latest mainconcept or CCE.

You seem to be inconsistent in your advice, or you didn't make yourself very clear. On other posts you said that in TMPGenc CQ_VBR is better than 2 pass VBR. Also, if I remember correctly, you said that TMPGenc 2 pass VBR is better than both Mainconcept and CCE. So, what is your opinion here?

[ChrO]

Thank you for the replys!
I know I should add the transitions before but i had allready encoded them to mpeg when i decided to add them, and tmpgenc takes time...
And EVE only encodes what is neccasary.

I will try to use 2 pass encoding, with the same clips, and the same max/min bitrates. But i thought that tmpgenc would take the bitrate neccasary for CQ 75 no matter how much is needed?

It has nothing to do with the max bitrate of 8000, thats enough?

Thanks again!
ChrO

[SatStorm]

I always refer to CQ mode, which covers both CQ or CQ_VBR.
And a more diplomatic way to say "equal" or "near same" quality is "it don't worth the time"
You know the lines of the judgments on this matters is more of a personal taste and objervation. For the average user, is not a bid deal afterall. Nor someone with a mainstream TV.
Anyway, just because you ask to clearify:

IMHO, on the latest TMPGenc versions, 2 Pass VBR only serve the need to know a specific filesize, before the proccess. In quality terms, CQ does the same, and some time better, in almost half time!

If you do some tests with latest TMPGenc, you gonna see that if you compare 2 Pass VBR with CQ mode, you end up with picture quality almost identical. Sometimes CQ look better, sometimes 2 pass look better, with the difference never be actually more than 3 - 5 % overall on quality.
In my eyes, CQ look better on static and really fast scenes, while 2 pass VBR on average moving but colourfull scenes.
In short terms, the difference between the two (CQ - 2 Pass) is almost unoticable. The only fact and certain thing between the two, is that CQ encoded files are always smaller! Expecially if you filter your source (a VHS capture for example)
But this talk here is what I call ironicly "TMPGenc-ology". All the other encoders act really different on all this, and follow what the book says. And the book says that 2 Pass VBR is always better 1 Pass VBR, if the average bitrate is the same. A comparison really difficult to do in realife, because how you can force a CQ encoding act 100% average base? It is impossible!

About mainconcept's 2 pass VBR encoding, an encoder I testing extremelly the last 2 weeks, I believe that:

Mainconcept smooths the picture, while TMPGEnc have it sharper. But:
On static backgrounds, Mainconcept do a better job than TMPGenc: There are some pixelaration, blockness if you like to call with TMPGenc, which is unexplained. Those don't exist with mainconcept! Now, for cartoons, this is a huge issue you know...

In praxis, if you wish to ecnode to 352 x 576 with TMPGenc, you have to rise the bitrate up to 2500kb/s (minimum value) when you use 2 Pass VBR just to avoid this strange effect. I'm talking about those colour blocks in the static background.
For reasons I can't explain, my tests show me that if I set CQ mode with a minimum of 1800, I have less of this block noise than when I set 2 pass VBR with a minimum of 2500 ! And I keep roling tests on this, just to find out WHY this is happening. I mean, I already eliminate the blocks you can see on TMPGenc with fade in/outs, after the correct resizing method issue, we already discuss before. But those blocks on the static backgrounds is a whole other TMPGenc issue, which probably bothers only me and few others, who over the years turn really picky on picture quality terms....

Anyway, the suprice is that you don't have this issue with mainconcept. This encoder don't create those blocks on static scenes with low bitrates.

So, you have sharpness & overall a better picture (TMPGenc) vs block free and smoother picture ( Mainconcept) when you use 2 pass VBR on those 2 encoders. It is more a matter of taste what you like or not.

And since bugster advice ChrO to try 2 Pass VBR, I point him to try it with an other encoder, because with TMPGenc probably won't see any difference compare TMPGenc's CQ.

For CQ vs CQ_VBR now, well... CQ looks to work better overall. CQ_VBR tries to set automatic a VBR value, in a % way, while CQ mode do it in a "rise over minimum" way. So IMHO CQ is more upredictable, but has more chances to produce better results. It doesn't try to determine something like CQ_VBR, just rise the bitrate anytime the encoder request for higher one.
I don't remember myself pointing someone to CQ_VBR. I mostly write "CQ (1Pass VBR)" when I refer to that kind of encodings, just to point that CQ mode is a VBR mode.
Do you remember where I pointed for CQ_VBR? The last time I refered to this mode, was almost 2 years ago (the famous SxVCD years!)

[W_Eagle]

I don't remember myself pointing someone to CQ_VBR. I mostly write "CQ (1Pass VBR)" when I refer to that kind of encodings, just to point that CQ mode is a VBR mode.

Aha, now I see where I got confused. Anyhow, thank you very much for your detailed reply.

[ChrO]

Hi again
Okay, now i have tested both mainconcepts and Tmpgenc, with exactly the same captured cartoons, in both 1 pass and 2 pass mode... And Satstorm you are absolutely right
I dont see any viewable blocks in mainconcepts encoder, but the picture is, like you said, a little more blurred. It is almost unnoticable on the tv screen.
In Tmpgenc, if i increase the min bitrate to 3000, Tmpgenc do not show the blocks!

So conclusion... In Tmpgenc, increase the min bitrate, and get the sharpest picture with no risk of blocknoise!

I could not tell the difference between 1 or 2 pass mode, if i gave 2 pass the same average bitrate as 1 pass, but that is probably logic.
Thanks very much!
Chr