Problem With Resolution Settings & MPEG Encoding

[Zeek]

Hello...

I have a fair amount of experience with MPEG encoders, but recently have come up with a new problem I'm hoping someone else might shed some light on. After reading up on things a bit, I've decided to modify the encoding process I use with VHS caps/acquisitions by changing the resolution of the final product from full D1 to half D1, or 352x480. Seems I've been wasting a lot of time and disk space by using full D1 with a source that is only about half that. Shame on me, but better late with the fix than never, I guess...

At any rate, for the most part, I use a camcorder with passthrough features to create an avi file, filter if needed, encode from there, and voila. My question pertains to how I should go about creating a resolution *other than* the standard 720x480 with these encodes. I've tried a few different encoders and, while the settings seem easy enough to navigate and tweak, the results aren't making sense to me. The details:
___________________________________________

Procoder 2:

Width/height settings are easy enough to find and change, but results achieved with the output files are kind of baffling. I've experimented with encoding at both 720x480 and 352x480 on a small file, just to have a look before moving on to the real thing - and, although the encoding time varies substantially, the size and appearance for both output files are the same! Similar appearance isn't really a surprise - but file size?!? That doesn't make sense. Am I doing something wrong here?
___________________________________________

TMPGEnc:

Under the video settings tab, some settings are highlighted (white/tweakable) and others are not. Unfortunately, the size is grayed out and set permanently at 720x480. There must be a way to change this, but I've yet to find it. Anyone know what it is?
__________________________________________

Mainconcept:

As with Procoder, settings are easy to find and change - but encoding at full D1 vs. 1/2 D1 yields almost no change in size of the output file.
____________________________________________


And so, my questions:

1. Am I missing something here with regard to resolution settings?

2. Is it not correct that changing resolution from 720x480 to 352x480 should reduce the output file size? Seems fairly obvious that it should - but again, if I am missing something here, please point it out to me.

3. Off topic a bit, but assuming I get to the bottom of this, how will the lesser resolution settings affect bitrate calculators? The videohelp calculator is pretty much all I've ever had to use - but its settings vary only by the *type* of movie (DVD, VCD, SVCD, etc.) and not the resolution. In other words, when creating a DVD with a 1/2 D1 resolution, since the lesser resolution should require less real estate than that needed for a "normal" DVD, how would I figure out my bitrate limits with the different resolution?

Thanks in advance to anyone that can shed some light on this for me.

Regards,

- Zeek

[edDV]

Resolution settings are independent of bitrate settings. Experiment with both. Of course you can get away with less bitrate with 352x480/576 resolution.

[daamon]

TMPGEnc: Under the video settings tab, some settings are highlighted (white/tweakable) and others are not. Unfortunately, the size is grayed out and set permanently at 720x480. There must be a way to change this, but I've yet to find it. Anyone know what it is?

In one of the sub-directories (i.e. below where the templates for different encoding templates are) there should be called "unlock" - use that and you can access previously unavailable settings.

1. Am I missing something here with regard to resolution settings?

Yes - The filesize of an encode is purely reliant upon running time and bitrate. So, encoding the same bit of footage at D1 and half-D1, if using the same bitrate will yield very similar filesizes.

2. Is it not correct that changing resolution from 720x480 to 352x480 should reduce the output file size? Seems fairly obvious that it should - but again, if I am missing something here, please point it out to me.

No - See answer above.

An encode at full-D1 will require a minimum bitrate before you start noticing quality loss. By reducing the resolution to half-D1, in theory, you'll need around roughly half the bitrate before you start noticing a similar quality loss. The point at which this happens is subjective and so can't be calculated.

However, because you have a lower resolution you don't need as high a bitrate to get a decent (acceptable to you) picture - this is where the smaller filesize comes in from using the lower bitrate.

3. Off topic a bit, but assuming I get to the bottom of this, how will the lesser resolution settings affect bitrate calculators?

It won't - bitrate calculators aren't affected by resolution. Only running time and end filesize. You need to find what minimum bitrate is acceptable to you for full-D1 and half-D1. And, even then, this will only be a guide coz different footage will have different bitrate needs.

In other words, when creating a DVD with a 1/2 D1 resolution, since the lesser resolution should require less real estate than that needed for a "normal" DVD, how would I figure out my bitrate limits with the different resolution?

Trial and error...

I hope that helps.

[Zeek]

Daamon,

It does help - a lot.

Your explanation wrt lesser resolution requiring lesser bandwidth to achieve the same quality makes a lot of sense. I usually try to keep my 720x480 stuff at around a 5-6 MB/S minimum. I'll give the half D1 stuff a try with a 3-4 minimum and see how it goes. With that out of the way, I can now get back to work on all of it.

Thanks a bunch...

Z

[daamon]

Pleasure. Glad to be of help.

Good luck.

[ranma_st]

Ok! I understand that the file size does not depend on the resolution. That with lower resolutions you need less bitrate to obtain similar quality.

However, what about if your capture is 352 x 480 and you encode it with half D1. If you use the same bitrate (let's say 4000) encoding in D1 and in half D1, the file sizes will be the same. However, the quality in the half-D1 wouldn't be better since you don't have to resize the video?

I would really appreciate you help with this, since I am thinking to change my VHS captures from D1 encoding to half-D1 encoding to get better quality with the same file sizes.

THX

[Zeek]

ranma_st:

You've got the hard part right, but you're confusing resolution and encoding a bit. D1 or half-D1 refers to the resolution itself, and has nothing to do with the encoding process whatsoever *other than* the fact that higher resolution will require a greater bitrate to look good. If you encode using a 4000ish (+/-) MB/s bitrate with half-D1 resolution, that should be just about right. Double that to 8000 and you'll have a bloated file that is carrying more information than can be utilized/displayed properly *at that resolution*. Cut it in half to 2000, and you'll have the opposite problem, meaning your picture will look like hell.

Lord Smurf has a graph on his website (digitalfaq.com) that is an excellent & quick reference to give you an idea of how much bitrate you need to allocate to a movie, depending on the resolution you choose. Also, do a search for bitrate allocation and that will get you started in the right direction.

Lastly, to clarify, "Full D1" resolution means simply that the resolution is set to 720x480 (DVD standard). Half D1 means 352x480 and, I believe, is the resolution that normal VHS uses.

Good Luck...

- Zeek

[ranma_st]

Zeek:

Thank you very much for your help.

Actually, I think I messed my question. I understand (a little) the bitrate part. Also, the D1 and half-D1 resolution.

What I really want to know is: Is there any gain when you encode a 352x480 captured file in half-D1 mode due to the resize (or lack of) during the encoding?

I ask this because if you encode the same video (352x480) in D1, TMPGenc will have to resize the video (which I have heard TMPGenc is not good at). However, if you encode the 352x480 video in half-D1 there shouldn’t be any resizing (or should be?), thus creating a file with better quality (???).

I hope this makes my question more understandable.

THX

[jman98]

Lastly, to clarify, "Full D1" resolution means simply that the resolution is set to 720x480 (DVD standard). Half D1 means 352x480 and, I believe, is the resolution that normal VHS uses.

Good Luck...

- Zeek

Actually it's worse. According to Wikipedia, the resolution of VHS is 240x480 (NTSC) or 240x576 (PAL). So while VHS does have the same vertical resolution as half D1, it's worse on horizontal resolution. This is not even touching the speed of the original VHS tape which also plays a role in the quality of any transfers.

[Zeek]

Thanks for clarifying. I was wondering about that myself, just after I sent it. My bad.

I guess we can agree that VHS is pretty much sh*t quality, though.

[Zeek]

Sorry, ranma_st, I missed your post.

I'm also not sure that I'm quite understanding your question.

If you capture a file at 352x480 and then encode at half D1, there will be no resizing because half D1 *is* 352x480. If you encode that same file to 720x480, of course it is true that the encoder (TMPGEnc, or whichever one you end up using) will have to resize the file. There is no benefit, in terms of quality, to doing this.

Also, understand that ANYTIME you encode, you lose information. Ideally, the loss of information will occur in a spectrum that is not perceptible to us humans. This depends mainly on the quality and settings of the encoder. That said, a general rule of thumb is that encoding means a loss in quality.

If you're capturing VHS (my guess is that you might be), my recommendation would be to capture at 352x480, and then encode at 352x480 with variable bitrate set to an average of around 4000, maximum at 5000 and minimum at maybe 3000 or so. Hope this helps.

[edDV]

Lastly, to clarify, "Full D1" resolution means simply that the resolution is set to 720x480 (DVD standard). Half D1 means 352x480 and, I believe, is the resolution that normal VHS uses.

Good Luck...

- Zeek

Actually it's worse. According to Wikipedia, the resolution of VHS is 240x480 (NTSC) or 240x576 (PAL). So while VHS does have the same vertical resolution as half D1, it's worse on horizontal resolution. This is not even touching the speed of the original VHS tape which also plays a role in the quality of any transfers.

Since VHS is analog in the horizontal direction, it has no pixels, only analog bandwidth which in the case of VHS is maximum 3MHz for luminance. The vertical direction is sampled by the scanning process so there are 480 actual lines for NTSC VHS, (576 for PAL).

3MHz luminance translates into about 240 horizontal "lines of resolution" which means an EIA resolution test chart's alternating white-black lines blur at 240 lines in the horizontal direction. This does not mean 240 pixels. A rule of thumb adds a 1.2 to 1.4x multiplier to get the equivalent pixel count or ~288-336x480/576 to see "240 lines of resolution".

NTSC broadcast maxes at 330 lines of resolution. To get everything, a pixel sampling of 400-460x480 would be needed. 352x480 falls between VHS and NTSC broadcast.

Cable "Digital" channels use ~ 524x480 for MPeg2 sampling. Satellite uses 480-524x480 MPeg2.

EIA test Chart.

[ranma_st]

Ha, ha!

guess we can agree that VHS is pretty much sh*t quality, though.

I am not sure if 20 years ago someone would have said something like that.

Anyhow, I have search and read in this site a lot looking for an answer to this question. I don't know but I think that some of my quality problems when converting a captured file to DVD have something to do with the resizing in TMPGenc. Then, I read about half-D1 and I want to know if that's a viable solution or I am just speculating.

I seldom post here, since I always find answer to my questions searching the site. However, this one seems to be very elusive.

[ranma_st]

Zeek:

Thank you for your time.

If you capture a file at 352x480 and then encode at half D1, there will be no resizing because half D1 *is* 352x480.

Exactly what I thought. What I am currently doing is capturing at 352 x 480 and encoding to DVD D1 (720 x 480). Since TMPGenc has to resize the video, I thought that there should be some quality loss. So, of course no benefit in doing it.

What I wanted to do is what you just recommend, encode to half-D1 (352x480). However, I first wanted to know if there would be a real benefit when doing it. I know that the first benefit will be the bitrate allocation (if I use the same bitrate I use in D1). Then, I wanted to know if by avoiding the resize part of the encoding, I will get more benefits.

I think you answered great part of those. However, any additional comment will be more than welcomed.

[edDV]

Ha, ha!

guess we can agree that VHS is pretty much sh*t quality, though.

I am not sure if 20 years ago someone would have said something like that.

Anyhow, I have search and read in this site a lot looking for an answer to this question. I don't know but I think that some of my quality problems when converting a captured file to DVD have something to do with the resizing in TMPGenc. Then, I read about half-D1 and I want to know if that's a viable solution or I am just speculating.

I seldom post here, since I always find answer to my questions searching the site. However, this one seems to be very elusive.

In theory 352x480 is in the ballpark for encoding VHS but purists will use 720x480 to be sure. In any case, if you are going to filter the image it is better to capture at 720x480, filter (e.g. levels correct, resize, noise reduce IVTC, etc.) and then encode to 352x480.

You should try all 3 ways on a test clip sequence and see which you prefer with your capture and display hardware. Mileage will vary as they say.

1. 720x480/576 capture - 720x480/576 encode
2. 720x480/576 capture - 352x480/576 encode
3. 352x480/576 capture - 352x480/576 encode

[Cornucopia]

Couple of things:

D1 is more truly 704x480 (that's why 1/2 of it is 352x480) than 720, but that's splitting hairs.

Yes, you "lose" something when you have to resize--No matter what. Why? Because resizing requires mapping old pixels to new when it's not 1:1, and that usually means guessing (interpolation). Whether UP or DOWN.

Resizing DOWN is almost always better in terms of maintaining the perceived quality. Has to do with the similarity of this function with OVERSAMPLING.

So, it's ok to cap at D1 and then resize down, or to cap at 1/2 D1 and keep at that size. Avoid the resize UP.

Note: If you're using Analog passthrough, that's a DV stream--which should always be left/saved at 720x480(576). If you're editing + processing/filtering, the rendered file can be saved as 1/2 D1 then (preferably using None/Lossless codec) for later encoding.

Of course, 1/2 D1 is RESIZED UP in the DVD player hardware, so you need to do some end-to-end A/B/C comparisons to see which gives you the best overall look for your material.

Scott

[Zeek]

ranma_st,

Glad to be of help. I'm going to defer to the guys that posted the last two answers here on the rest, because they've made some very good points. And, for what it's worth, I myself do my VHS caps at 720x480 and then scale down to 352x480 when encoding. I usually use the camcorder passthrough feature that Cornucopia is talking about - and it works very well.

My advice is to take edDV's advice and run some small trials to see what you like best. Video is still an "art" as much as it is a science and, at the end of the day, what really matters is what looks good to you. That said, if you are capturing VHS and making it DVD, it's pretty hard to go wrong by capturing at 720x480 and then encoding to half D1. Run a few trials though, and see what you see!

Z