Poor quality transition fades on mpeg-2

BruceDale wrote on 9/28/2010, 11:30 AM
I am squeezing 3 hours of video on one DVD. I am rendering as MainConcept MPEG-2 for use in DVD Architect. I'm using 2-pass Variable bit rate at 3,000 kbps which is as low as I've ever done. However, I have done it before, a year ago, and the fades were fine. Now the fades out to black and back in again are extremely boxy. That is, the video shows boxes all over the place that change frame by frame. Does anyone have any ideas how to make the movie render better?

Comments

Former user wrote on 9/28/2010, 11:59 AM
VBR will cause this more than CBR. But at that low a bitrate, you are going to see any fades and dissolves. The Mainconcept encoder does not handle fades very well, in my opinion.

Dave T2
musicvid10 wrote on 9/28/2010, 1:52 PM
You need a VBR minimum of 1,500,000 to minimize blocking in fades and transitions. The default of 192,000 is woefully inadequate. If your VBR average is 3,000,000 then there is not much reason to use VBR -- use CBR instead.

Personally I can't stand anything below 4Mbs on a DVD -- unless it's a football game where I have no choice.
farss wrote on 9/28/2010, 1:57 PM
1) What are your Min and Max encoding settings?
2) You could try putting a marker at the start of the fade to force an I frame.
3) You could try something like Mike Crash's Dynamic Noise Reduction filter. Reducing noise before the video reaches the encoder should help at low bitrates.
4) Combine keyframed Gaussian Blur with the fade to / from black.

Bob.
Chienworks wrote on 9/28/2010, 3:12 PM
"If your VBR average is 3,000,000 then there is not much reason to use VBR -- use CBR instead."

I'm not sure i follow this. VBR is precisely intended for lower bitrate situations. It allows bits to be "saved" up to be used when they're needed. If the average bitrate approaches the maximum then there's no benefit with VBR, but when the average is a lot lower then this is exactly when VBR is the most useful.

I do agree though that results can be improved by raising the minimum. I never use anything below 1,000,000.
john_dennis wrote on 9/28/2010, 3:25 PM
Christmas is coming. You could start asking Santa for a Blu-ray writer or a 100 piece spindle of Verbatim dual layer disks and use more than one. Unless, of course, you are making a DVD for someone else. Then, do what they ask and take the money. 3 mbps is a mighty low bitrate. It seems you are putting yourself through a lot of trouble for the sake of using one DVD.
TheHappyFriar wrote on 9/28/2010, 3:39 PM
Couldn't he render out his transitions to a high-bitrate mpeg-2, render the rest at VBR & then assemble them in a new Vegas project & render out to VBR & have it smart-render everything?
musicvid10 wrote on 9/28/2010, 4:10 PM
I'm not sure i follow this.

Good question Kelly; I've come to the conclusion that there has to be adequate range between min-av-max bitrates for VBR to offer much of an advantage.

At such low ABR, there is not much wiggle room for the variable bitrate to do its work. Even with a high maximum bitrate, the advantage is negligible because of the small crawlspace between an adequate minimum (1.5Mbs for me) and the OP's choice of 3Mbs ABR.

It is precisely the ability to go far below the average when it would be tolerable that gives VBR the ability to go far above average when it would be advantageous. Another way to say this is the amount of space between the min-average bitrates limits the time VBR can spend working much above the average bitrate.

The extreme test of this theory is that if the minimum and average bitrates were both set at 2.5Mbs and the the max at 9.5Mbs. The result would be the exact equivalent of 2.5Mbs CBR.

Tests I ran several years ago showed for my purposes (ymmv), that:
1) 1,500,000 min is my absolute bottom, and 2,000,000 min is visibly better when there are FTB transitions,
2) At least 2Mbs "windows" (preferably 2.5+) between min-average and average-max bitrates are necessary for VBR to show an advantage that I can appreciate, render times notwithstanding.
3) An ABR of 4Mbs under any circumstances is about the absolute worst I can stand.

That being said, sometime this fall I'll do some mpg test renders at 1.5 / 3 / 9 VBR, compare and post back my impressions using both high detail and high motion. I'm always open to having my assumptions challenged. That I'd ever use 3Mbs except for that occasional AFC playoff game is still pretty unlikely.
farss wrote on 9/28/2010, 4:49 PM
Kelly is correct. To take it one step further there's only a weak correlation between bitrate and image quality. As pointed out recently you can encode images of pristine quality at almost 0Mbps, just so long as nothing changes between frames i.e. a slideshow.
The correlation between source video quality and rendered output quality is very strong. We have a set of commercal DVDs that have 6 hours of video on single layer DVDs. All studio content, well lit and shot with good cameras. Big areas of the frame with the same content e.g. blue sky, clean backdrops etc help the DCT compression. Low noise, locked off shots, smooth slow pans etc help with the temporal phase of mpeg-2. Put all of that together and you can see why Hollywood DVDs look so good.

As suggested hand encoding is one way around the problem at hand, wrangling the content is another and if you have deep pockets then a better encoder that does more than two passes is another solution. Also worth considering dropping the Quality slider a tad at low bitrates.

Bob.
musicvid10 wrote on 9/28/2010, 4:57 PM
"Couldn't he render out his transitions to a high-bitrate mpeg-2, render the rest at VBR & then assemble them in a new Vegas project & render out to VBR & have it smart-render everything? "

I don't know if this would work exactly according to plan in Vegas, but the principle is totally sound.

All of my demo DVDs have chapter titles with bright colors, sometimes motion, and always FTB. I always render them at 8Mbs CBR and stitch them together with the program segments in VideoRedo before sending the whole mpg over to DVD Architect. Doesn't take as much time as it sounds like.

Would you have time to test and develop this technique in Vegas and report back to us?
rs170a wrote on 9/28/2010, 7:58 PM
I know I've mentioned this before and now seems like a good time to do it again.
I shot a play earlier this year that ran 2 hr. 30 min.
I was positive that, due to it's length, it would have to be either a dual-layer burn or 2 DVDs.
Out of curiosity, I did a short test and the result amazed me.
I ended up burning the entire play on a single layer disc with the following 2-pass VBR settings.
2,224,000 / 3,712,000 / 6,496,000
I had the advantage of shooting it with our JVC 550U (1/2" - 3 CCDs) equipped with a good lens (i.e.. not the stock one) and the show was professionally lit with very few extremes in lighting conditions.

Mike