Tried Procoder 1.5 and it produced better output than CCE...

[Matthew]

Had a noisy DVD source (PAL mixed interlaced/progressive, 16:9, 1.78 aspect ratio) and re-encoding at a bitrate a little over 3100, which ain't great. I tried CCE, and then ProCoder.

Here's the same frame, from original, CCE and procoder. It's typical of a lot of frames.

http://members.tripod.com/auntyjacksret/procoder/1-original.jpg
http://members.tripod.com/auntyjacksret/procoder/2-cce.jpg
http://members.tripod.com/auntyjacksret/procoder/3-procoder.jpg

As you can see there's less noise in the ProCoder encode than CCE. They may achieve it by blurring things a little around edges, I don't know, but that's better than haze and better than using an avisynth filter than blurs the whole frame.

As an aside, Photoshop assigned the following sizes to the jpegs, which were all saved in the same manner:

71212 (original)
71119 (CCE)
63837 (procoder)

Maybe that gives credence to the theory that ProCoder blurs the video...

[bb]

Matthew,

comparing single frames can be misleading. I'd like to know if the pictures show I-frames for both CCE and Procoder.

By the way: to view the pics you have to visit http://members.tripod.com first, then you can proceed to the the picture link (no direct linking possible).

bb

[Matthew]

I agree showing single frames can be misleading (differing settings can also be a problem), it was just one I grabbed to demonstrate what is happening in many frames. And something that is consistent with what others have said before.

I don't know if either is an i-frame, the odds are against it I think.

If you are interested I can provide a short clip, e.g. just trim out the first 30 secs (say) of each encode using M2Edit Pro.

Thanks for the tripod tip, they are a real pain in the backside.

[ArchStanton]

Well, I think it´s generally accepted that procoder 1.5 produces better results with inerlaced material than CCE.

And you´re right, from what I´ve read, the better quality is a result of internal filtering in procoder.

I also have a little question, is version 1.5 still so painly slow as its predecessors?

[Matthew]

It seems to be better for the progressive frames as well though, the frame I posted looks progressive for example. So perhaps it's more about whether the material is noisy (such as DV).

Incidentally I'm doing only frame based encoding, I don't want problems in standalones.

Procoder in mastering mode isn't the fastest, but it varies a bit. I'm encoding eps which are almost an hour and they took close to 2 hours each with CCE 4 pass VBR, but with ProCoder they seem to take more like 2.5 hours. However the ep I took the sample frame from took close to 3.5 hours. Perhaps the internal filtering was having trouble with the extra noisyness of that ep.

[Yuri]

Matthew, HI!
Unfortunately, couldn't see your frames because of the site,
such a shame:(

[Kedirekin]

It's the same old anti-leach referrer page trick - as long as you start from any tripod page it will let you through.

Do view source in your browser. Find the URL for the link in Matthew's post. Copy it to the clip board.

Click the link. When you get the tripod page, paste the URL into the browser address bar (CTRL-V to paste the URL from the clipboard). Then you can get to the image.

[bb]

Originally posted by Matthew
If you are interested I can provide a short clip, e.g. just trim out the first 30 secs (say) of each encode using M2Edit Pro.
If it's not too much effort for you, I'd be interested :)

bb

[Matthew]

For size reasons I made the sample only 100 frames, and am uploading the original video, cce video and procoder video for this. I'll PM you with the address.

It's candles against a black background (maybe not the best, I just used the first 100 frames), but if you look at frame 38 (for example) in virtualdubmod, you should see a clear difference in the level of noise.

Don't trust the last couple of frames as womble mpeg editor re-encoded them (and boy is it bad at it).

[bb]

Thank you very much for the clips. I used the following AviSynth script to compare your clips side by side in VirtualDub while stepping through the frames:

v1=MPEG2Source("C:\Dokumente und Einstellungen\All Users\Dokumente\Downloads\Video\Test Clips\orig.d2v")
v2=MPEG2Source("C:\Dokumente und Einstellungen\All Users\Dokumente\Downloads\Video\Test Clips\cce.d2v")
v3=MPEG2Source("C:\Dokumente und Einstellungen\All Users\Dokumente\Downloads\Video\Test Clips\pro.d2v")

tr=24
cx=180

v1=v1.Trim(tr,0)
v2=v2.Trim(tr,0)
v3=v3.Trim(tr,0)

v1=v1.crop(cx,0,-cx,0)
v2=v2.crop(cx,0,-cx,0)
v3=v3.crop(cx,0,-cx,0)

StackHorizontal(StackHorizontal(v1, v2), v3)


The Procoder clip's quality is clearly superior in your test clip. Which CCE version and filter settings did you use in your test encode? Which were your Procoder settings?

So far I guess you can only make an assumption about the resulting quality using the encoder's standard settings. Furthermore I believe that CCE will perform better if you let it blur some more before the encoding step. In CCE you can control that (at least in V2.66+).

bb

[Yuri]

Thank you most kindly, Kedirekin.

[Matthew]

Thanks for taking a look bb :)

Originally posted by bb
The Procoder clip's quality is clearly superior in your test clip. Which CCE version and filter settings did you use in your test encode? Which were your Procoder settings?

I used 2.67.0.11 with:
-4 pass VBR
-GOP size of 12 (3/4) - perhaps 3/5 would have been better because of the low bitrate relative to nature of source.
-All filters disabled
-Flat part priority 32 - probably should have been lower
-Intra dc 10 - 9 may have been better, for same reason as with GOP size.
-Progressive unchecked
-zigzag scanning order - I used zigzag because I originally thought this ep was entirely film, but on closer inspection it turns out there is interlaced content in there. But less prevalent than in later episodes (there are 15 in all).
-interlaced=false in avisynth script for same reason as used zigzag

ProCoder settings:
-GOP size 15
-Intra DC 9
-Interlaced, upper field first.
-No option for zigzag/alternate, this is automatically selected on a frame by frame basis, says Canopus.
-2 pass VBR, mastering quality.
-interlaced=true in avisynth script because I realised there was some interlaced material and apparently the wrong setting on interlaced material is much more damaging than the wrong setting on progressive content.

Bitrate was 3170 and max 9200 (quite arbitrary really).

Now, I know those CCE settings may not have been optimal or consistent with ProCoder's, but I think that reducing the flat part priority setting by 10 (say) isn't going to make much difference.


Originally posted by bb
Furthermore I believe that CCE will perform better if you let it blur some more before the encoding step. In CCE you can control that (at least in V2.66+).

I take it you mean just better than it had been rather than necessarily better than ProCoder? What setting to enable and is it "global" like avisynth's blur feature? It's one thing to blur around the edges but to blur the whole picture is another matter (although perhaps I am wrong to think ProCoder only blurs around the edges in order to reduce noise).

Incidentally, ProCoder supports direct input of M2V source but I tried it and the length was detected wrong (2 secs short). I take it there'd be no quality advantage to using ProCoder to decode the stream rather than DVD2AVIU/mpeg2dec/avisynth? All frames seemed to be marked as interlaced/zigzag (I checked a couple of i-frames in an interlaced part using VobEdit) so access to those flags on a frame by frame basis mean nothing.

[magilvia]

It seems to me that in these shots CCE had done a really better job in reproducing the original image. The original was noisy and the resulting CCE image is noisy.
If an image is more similar to the original I think it is of better quality.
Is it true that Procoder gives a better-looking image but it achive sthis while altering the original. I don't really like this.
Anyway you should try to activate the anti-noise filter in CCE and reencode the entire movie. Pheraps this will lead to an even better looking image.

[bb]

Matthew: I don't know if CCE or Procoder will produce a better video from your test material if either one is tweaked to the maximum for quality.

But one thing is for sure: Your test settings were much too different to make any assumption. You have to redo the test with settings as similar as possible.

bb

[Matthew]

/me groans

I'll just re-encode that short clip a bit with CCE, that won't be too much effort. However, I have a feeling that I'll not get output better than ProCoder. It seems as if Canopus has really focused on getting good output for noisy material, hence people like DDogg (a prolific tester from everything I've seen) swearing by ProCoder for DV (well at least he did, anyway).

[Matthew]

Well I re-encoded that clip with a bitrate of 1800, which gives slightly larger size than CCE gave to original encode (and this is more than ProCoder allocated, 1500 in CCE would give equivalent size). I did it firstly with the original settings and it seemed to look the same (i.e. relatively bad).

Then I changed the settings to GOP 3/5, intra dc 9 and FPP of 16 (default) and the noise was reduced. I ended "effect restricted vertical filter" (anti-noise) and set it at 16, the max recommended for interlaced material (this is a mixed video stream after all) and the improvement compared to the original encode was significant.

If you want to have a look [please =)] I uploaded this last sample to the same site as before, in the link just use the filename cce2.rar instead of sample.rar.

Personally I'm not so sure ProCoder is better now, of course that's based on this sample alone, what the lower FPP and the anti-noise filter will do to other parts of the picture in lighter coloured scenes I don't know. And the CCE's manual about the anti-noise filter makes me nervous: "To use this filter on interlaced material, specify 16 or less, otherwise, a movement will be jerky."

Finally of course there is still the issue of whether ProCoder will handle the interlaced parts better as many people have claimed.

[fungus]

I have been using CCE 2.5 for quite a while now within DVD2SVCD, when the chance to use Procoder 1.5 arose. In comparison I found that it was slow by comparison. I ran the encoding cycle with both CCE and Procoder on the same files and faced a 12 hour wait over the 'normal' 6 hours I normally have. The quality differences were minimal and did not justify, IMHO, waiting twice as long for the end result.

However Procoder comes into its own when 'stitching' bits of film together.:)

[DSP8000]

I've asked a question about Canopus Let's Edit editing capabilities & gues what? I ended up buying it + Canopus Procoder Express.
Maybe I did that because I can't wait for ProCoder v.2 but,looks like Canopus Procoder Express is using the same encoding engine as the new ProCoder & almost the same improvements.
Recently,inspired by this thread I did backup of Titanic on DVD-RW with Procoder Express & I honestly got supprised by the quality.
I have one backup of this movie done with CCE SP v.2.50+DVD2DVD-R but the quality it's not as good as the Express's output.There is one scene with big zoom on top of the boat at the begining & CCE output shows lots of artifacts & macroblocks but Express's output nothing,clear picture...amazing
I'll do DV video with the two encoders to see if there will be any difference.So far I'm more than happy with my package Canopus Let's Edit+Canopus Procoder Express.

DSP8000