Encoder/transcoder, bitrate/compression level

[Bruce Ballslap]

Hi!

I guess pretty much all of you got frustrated just by seeing the title :) Bare with me please, I believe this post is in place.

Having read through many video quality comparisons, I still haven't found answer to these considerations. The main reason being that most of the tests are done using very low bitrates.

Now, I have set a rule for myself of not transcoding when the resulting bitrate would be less than around 4.1 Mbps. The rest I split.

Question #1: How big are the encode/transcode quality differences between CCE 2.7 3-pass vs. Shrink 3.2 w/ deep analysis and AEC at resulting bitrate of 4.1Mbps?

Question #2: Do transcoders/encoders cope worse with higher compression rates even when the resulting bitrate is as high as 5 Mbps (compared to low compression rate & low end bitrate)? Example:

Which is more challenging for a. encoders b.transcoders?
Case C. 65% of original --> end bitrate 5 Mbps
Case D. 90% of original --> end bitrate 3 Mbps

I guess that in question #2 I'm trying to ask if compression level affects the end product quality even if the end bitrate is high.

I've learned that bitrate has the greatest affect on quality, but maybe dramatic bitrate drops "have their say"? (Again: 8500 Kbps --> 5000 Kbps vs. 4500 Kbps -->3500 Kbps).

I hope this question makes sense and that you find it relevant.

[jdobbs]

People are always looking for simple answers to complex questions. Unfortunately it is like asking "How many pancakes does it take to cover a dog house?" How big is a pancake? How big is the dog house?

Quality is made up of (at least) three components in this context. Bitrate, compression level, and complexity of the source. You always have to consider all three. There are others, like effectiveness of the encoder/transcoder -- but I'll leave those aside right now. A non-complex source can be reproduced with a very low bitrate and a very high compression level and have little or no loss of quality. A very complex source requires a higher bitrate and a lower compression level. So there is no answer to your second question because it depends on what you are attempting to encode.

As for your first question... well again it isn't that simple.

First let me say that encoding (with a decent encoder) will always, always result in better quality than within-the-compressed-domain transcoding. End of story. Book closed. Fact.

Within-the-compressed-domain transcoders never get the opportunity to analyze the source picture. In fact they never see it at all. That's why they are so fast. They just remove motion vector and/or quantization detail from what was there in the original encode. The problem with that is that if the original was easy to encode, the preexisting bit distribution may not have been optimal -- and on the other end of the scale, if it was hard to encode, the output can suffer greatly from only minor changes.

If you transcode an easy-to-encode source at a low compression level and high bitrate -- it may look great and the difference between the transcoder and encoder may not be obvious. But it's there.

The times you really see the difference is when the bitrates, compression levels, and complexity make the job challenging. Then it is intuitively obvious to even the casual observer.

Also a lot of people don't realize that their equipment is "masking" errors. A poor quality television, poor connection, or low fidelity connection (composite for example) will create smoothing effects that cover errors. But when you finally upgrade (everyone does) -- you'll see them loud and clear.

It is possible to create a good backup with a transcoder. But since I know I will always get a better picture with an encoder -- why take the chance? I think its better to just do it right the first time. The only cost for the increased level of assuredness is a little extra computer time -- usually while I'm sleeping.

But then again, that's just my opinon.

[Sharc]

There is no absolute truth about quality, as quality is much related to your personal perception and expectation, and the bitrate required for providing a satisfactory quality depends strongly on the original footage (movie or anime, content of motion/action, degree of details etc.)

I try to give an answer, based on my experience:

Question 1:
At 4.1 Mbps average bit rate, I would - for typical movies - expect clearly superior quality from an encoder compared to a transcoder, irrespective of the degree (%) of compression.

Question 2:
Case C:
I bet an encoder gives better results compared to transcoding at 65% compression and an average bitrate of 5 Mbps. I would even expect excellent quality at 5 Mbps final bitrate from an encoder. Although the 5 Mbps may also be reasonably comfortable for Shrink, it is likely to exhibit stronger "pumping" for example (a kind of fluctuation between frames) at 65% compression, and possibly macroblocking and artefacts at fast action scenes and/or scene transitions, for example.

Case D:
At 3 Mbps you fall short of bitrate anyway for a typical movie. Quality may not be too exciting (but still acceptable) with an encoder (Some encoders are reported to perform well even at low bitrates down to 2.5 Mbps). I doubt however whether you will get anything acceptable with transcoding (Shrink) at such low average bitrates, irrespective of the % of compression.

In conclusion, final bitrate is crucial to the quality rather than compression (%). If processing time is not an issue (why should it?) I would prefer encoding over transcoding in any case. This is not to say that DVD Shrink cannot do a reasonably good job in cases where final bitrate is sufficiently high, say above 5 Mbps, and you are watching the result on a (small) TV.

[TomBrooklyn]

Case D:
At 3 Mbps you fall short of bitrate anyway for a typical movie. Quality may not be too exciting (but still acceptable) with an encoder (Some encoders are reported to perform well even at low bitrates down to 2.5 Mbps).

Most of the movies I back up seem to wind up with bitrates around 3000 or slightly higher. (I usually keep extras like how the movie was made and dump trailers, and I almost always dump the premium Audio track on the main movie for necessity of bitrate considerations.) I usually use the HC encoder.

I've been quite satisfied with the result of most of the backups I've made watching them on a 32" conventional TV with component wiring--I've found most backups seem close or very close to the original except in a few scenes here and there.

However, I will be getting a larger HD TV in the future and I'm wondering if that will reveal a noticible drop off in quality between the original and the backup with bitrates around 3000. Thoughts?

[jdobbs]

I watch my DVDs on a 50" Sony LCD System. It is high-def, high-end, and is very unforgiving with inferior sources. I do all of my backups keeping extras. The only thing I remove are the non-English languages and DTS. Any imperfections in the backup will definitely show up on my system -- and, believe me, anything done with a transcoder is very, very obvious (pixelation and blocking).

I, quite honestly, can't tell the difference between the original and my DVD-RB backups on most discs. The rare exception may be an extra that was reduced too much (that's because many discs have the extras encoded at a borderline bitrate even on the original).

I usually use CCE for my backups.

[therat]

hi jdobbs, would you mind sharing what settings you use to do a full backup please especially in regard to the extras.

cheers

[jdobbs]

I use the "out of the box" standard settings for most.

[candsh]

DVD-RB and CCE Basic. All just like they came to me...never made a change, always perfect.
The Transcoders were faster but now with a dual core CPU it takes about an hour for most and at the very most about an hour and a half...I remember the last movie I did with Shrink, a couple of years ago took longer than that.
To really go back I remember using ReMpeg, IFO Edit and a couple of other programs to get a movie and took 8-12 hrs. and then you would have to remix them because the sound was off ......Phew! I figure that jdobbs will make the changes and send them to me with a new version and then my job is to let him know if something doesn't work......funny thing I have NEVER had to make that call. But I also realize that all you guys that love to tinker have made a better product for me to use, however I am still watching the very first back-up I made with RB and it still looks good!!

[jdobbs]

:) My heart has been officially warmed.

Hmmm. ReMPEG. The concept was really good, all the GOPs were kept identical to the original. Too bad it didn't use a better encoding engine. We've really come a long way since then, eh?