mpeg screen size issue converting AVI off miniDV to an MPEG

[steve_r]

Hi,

I have been converting AVI movie files captured from mini DV tapes to mpegs for a while now....and I use Adobe PRM with a video server plugin along with TMPGENC to get my mpegs.

The above has been working great for mpegs created in pixel size 352X240.

Now here is the problem.....I tried the same conversion but this time with a bigger pixel size for the mpegs.

Trying a screen size: 448X336

I get a filesize exactly the same mb size as the 352X240 but this time the 448X336 videos are fuzzy and the quality is poor.

What have I not changed in my settings to give the 448X336 mpegs the same quality that my 352X240 videos have?

The avi is always captured off the same camera and is always 720X480.

My tmpgenc settings are: 352X240, 29.97fps CBR 1150kpbs, layer2 44100hz 96kpbs.

and for the bigger mpeg that is fuzzy for some reason: 448X336, 29.97fps CBR 1150kpbs, layer2 44100hz 96kpbs.

Maybe I need to fix these options? As they work for 352X240 but not for 448X336 ?

Please someone help me out I am sure the answer is obvious but I am missing something here.

[junkmalle]

If you use a larger frame size you should use a higher bit rate.

Even for 352x240 1150 kbps is pretty much bottom of the barrel. 448x336 will need about 30 percent more bitrate. It will not be VCD compatible with that frame size and bit rate -- if that's your goal.

[jimmalenko]

The above has been working great for mpegs created in pixel size 352X240.

Now here is the problem.....I tried the same conversion but this time with a bigger pixel size for the mpegs.

Trying a screen size: 448X336

I get a filesize exactly the same mb size as the 352X240 but this time the 448X336 videos are fuzzy and the quality is poor.

What have I not changed in my settings to give the 448X336 mpegs the same quality that my 352X240 videos have?

One word - Bitrate.

The only two items that affect filesize are running time and bitrate. Since your running time is identical, and your bitrate is identical, therefore your filesize is identical. Frame size does not affect the filesize.

Now as for quality, bitrate and frame size go hand in hand here. You need to use a bitrate that can sufficiently describe any given second of video. Unfortunately there's no real hard and fast rule here, as the required bitrate can vary greatly depending on the complexity of motion in your capture, the amount of noise, shake (particularly if handheld), and probably a couple of others that escape me right now. If quality is your main aim, take a look at using CQ in TMPGEnc - set it at a value between 90 and 100 and make sure that your max and min are suitable - I'd suggest somewhere around 4000kbps should be sufficient for your max, and you can leave min set at 0.

[steve_r]

Very helpful....whats the difference between CQ and what I was using which was CBR constant bit rate?

If I put my CBR from 1150 to 4000 will that help alot or should I just use CQ and do 0min 4000max and by value did you mean quality 0-100 and do 90-100 here?

what about P picture and B picture
P is 0 and B is 20 should I leave that like this?

VBV buffer size is also 40 is that good?

thanks for the help.

[jimmalenko]

Very helpful....whats the difference between CQ and what I was using which was CBR constant bit rate?

CQ is a one pass VBR - see https://www.videohelp.com/forum/viewtopic.php?p=1366332&highlight=#1366332 for links that discuss the whole CQ vs CBR vs 2-Pass VBR debate.

If I put my CBR from 1150 to 4000 will that help alot or should I just use CQ and do 0min 4000max and by value did you mean quality 0-100 and do 90-100 here?

Do both. Do a CBR @ 4000, and then do a CQ 0 min 4000 max and quality of somewhere between 90-100. Then compare. See which one you like quality-wise. You can use the source range function to isolate your footage down to a 1 minute sample also.

[junkmalle]

In short:

Constant bitrate uses the same number of bits to describe each frame. A completely black frame will take just as many bits as a very complex frame. A static sequence will take just as many bits as a very high action sequence. You know exactly how big the final file will be -- running time times bitrate.

Constant quality uses whatever bitrate is needed at each frame to get the desired quality. A black frame will take almost no bits to encode. A complex image will take a lot. A static sequence will take far fewer bits than a high action sequence. The final file size will be whatever is necessary to obtain the desired quality.

Variable bitrate lets you specify an average bitrate for an entire clip. This way you know how big the final file will be (average bitrate times running time again). The encoder uses two (or more) passes. During the first pass it uses a constant quality setting and keeps a record of how many bits each frame needed. During the second pass it allocates bits to each frame to meet the final average bitrate.

With both VBR and CQ you can specify an minimum and maximum bitrate. DVD players, for example, can't handle more than 10,000 kbps (total for audio and video) so you have to be sure not to exceed that. Some players also have trouble when the bitrate gets too low. So if a frame or scene requires more than the specified maximum bitrate it won't get it; if it requires less than the minimum it will be padded.

Note, although I refer to bitrates for "each frame" the encoder may not literally meet the requirements for every individual frame. The encoder may work at the GOP (Group of Frames) level or some other short interval.

[steve_r]

I tried both CBR at 4000 and 6000 kpbs
and it didn't really solve the quality problem.

I also tried CQ at 0-4000 and 0-6000 with 100% quality
and that didn't help either.

I am also getting like 3times the length in time of the actual movie I am encoding.

I did a 19 second sample and in 4000 kpbs its 57secs but only the first 19 plays than it stops.

At 6000 its 1min and 24 secs and only the first 19 secs plays.

What is going on here? How come it is doing this? Also maybe 6000 isnt even enough?

I am taking an AVI off MINI DV 720X480 and just trying to get a good mpeg-1 quality file at 448X336.

Not sure why this isn't working but I never had a problem doing this at 352X240.

thanks.

[junkmalle]

What exactly is wrong with the "quality" of the result? Maybe you could post a frame or a short clip. If you're happy with 352x240 at 1150 kbps and not happy with the same video at 448x336 6000 kbps there's something other than bitrate wrong with your encoding.

The problem with the length is usually a matter of variable bitrate audio in the source. TMPGEnc doesn't handle VBR audio very well. But DV AVI usually doesn't have VBR audio.

[midnightsun]

now excuse me a second here... if you encode 19 seconds of footage... how long do you expect the outcome to be?

if it takes 1 hour to encode it doesn't mean it's 1 hour long!

[steve_r]

Midnight the length problem has nothing to do with how long it takes to encode....but the actual run time of the movie.

When I did 1150 kpbs constant bitrate 19secs was actually 19secs when I played the movie in my WMP.

But 4000 was 57secs for the same 19sec file in WMP when I opened it.

And 6000 was 1min and 24secs for the same 19sec file in WMP.

The above is playtime not encoding time. which is weird and never happened to me before....everytime I made a movie the time I export is usually the time of the playback of the final movie. Strange I am not geting why this is happening.

------------------------------------------------

Here are all my samples and tests with the same movie.

First is 352X240 19secs playtime exported. MPEG 1, 29.97 fps, 1150 kpbs CBR, Audio layer 2 44100hz 96kpbs, open it in WMP and it's perfect shows 19secs total and plays all 19secs perfect.

THE ABOVE SHOWS PERFECT QUALITY. Just how I want it to be for the rest.

-----------------------------------------------

Second is 448X336 same 19secs playtime exported. MPEG 1, 29.97 fps, 4000 kpbs CBR, Audio layer 2 44100hz 96kpbs but for some reason when I open it in WMP, it's run time shows 57secs but it plays the actual 19 secs of the video exported than stops as if it were finished.

THIS ONE IS POOR AND FUZZY/BLURRY.

-----------------------------------------------

Third is 448X336 same 19secs playtime exported. MPEG 1, 29.97 fps, 6000 kpbs CBR, Audio layer 2 44100hz 96kpbs this time for some reason when I open it in WMP, it's run time shows 1min and 24secs but it plays the actual 19 secs of the video exported than stops as if it were finished.

THIS ONE IS POOR AND FUZZY/BLURRY.

-----------------------------------------------

Let me know what you think.....I tried CQ also with 0 min 4000 max got the same output as CBR at 4000kpbs...example 2 above.

[steve_r]

btw this is really pissing me off.....I am trying everything
even CQ with automatic VBR and still nothing works.

Maybe 448X336 just isnt meant to be?

For the screen size...the movie is great at 352X240

even if I double the screen to 200% it is great at 352X240.

But this 448X336 is a mess.

[jimmalenko]

Even though you should be able to do it, why is 448 x 336 so important ?

Why not stick to a "more standard" frame size ?


TBH I think it's definitely something you're doing wrong, as those settings I gave were very much on the conservative side.

[junkmalle]

Second is 448X336...
THIS ONE IS POOR AND FUZZY/BLURRY.

Something like the difference between this:



and this:



???

[steve_r]

jimmalenko it doesn''t have to be 448X336 can you suggest another size between 720X480 and 352X240 that might work well?

Junkmalle yep that is exactly it except the 352X240 is much better than yours there.

[jimmalenko]

If I'm not mistaken the 352 x 240 is deinterlaced, and the 448 x 336 is interlaced. View the 448 x 336 with software the deinterlaces properly like PowerDVD and I'm sure you'll find much ado about nothing. As for a frame size to recommend, why not stick the Full F1 (720 x 480) ?

[edDV]

Top frame has a blend de-interlace.

It isn't clear what you want to do with this 448x336 video. If it is intended for an interlaced DVD, that looks good to me. Author it, burn it, play it to a TV.

If you want it deinterlaced for some purpose, then deinterlace it.

[junkmalle]

Both of the images were made with TMPGEnc without telling it the source was interlaced. So yes, the top image is essentially blend deinterlaced because every scanline is now the average of pairs of scanelines. The bottom image isn't really interlaced anymore, it's an incorrectly resized interlaced image.

Steve_r -- I doubt you'll find any player that will play a 448x336 interlaced video properly. So if you want to use that size you'll need to deinterlace.

On TMPGEnc's Settings dialog, go to the Advanced tab. Make sure the Video Source Type is set to Interlace. The Field order is probably Bottom Field First. Down below check the Deinterlace box and select a deinterlace mode by double clicking on the line -- try Even Field Adaptation.

On the Video tab set the Encode mode to Non-interlace.

The best solution is to use a normal frame size like 720x480 and leave the video interlaced.

[steve_r]

Wow junkmalle that worked great...no more fuzz....now the 448X336 is the same size in mb as the 352X240 and both show perfect...at 1150kpbs.

But I am thinking maybe I should keep it at 352X240. or 720X40.

Btw these movies are not for DVDs or to play on any TV.

They are strictly for internet and mpeg to only play on a player like WMP.

[steve_r]

This de interlacing thing worked great junkmalle

thanks...
One question is there a down side to doing this? Like will the movie still be good with most players....same as the regular format I was encoding with before this?

thanks again

[junkmalle]

This de interlacing thing worked great junkmalle

thanks...
One question is there a down side to doing this? Like will the movie still be good with most players....same as the regular format I was encoding with before this?

thanks again

When you deinterlace you're throwing away some of the picture. You lose spacial and temporal details and you introduce some picture artifacts.

If you want odd sizes like 448x336, or if you want video that plays cleanly on all computers, you'll have to deinterlace.

If you're making DVDs stick with 352x480, 704x480 or 720x480 and leave the video interlaced. It will display properly on TV.

NTSC TV is broadcast as 60 half pictures per second. Each half picture contains alternating scanlines of the display, called fields. First the field containing all the odd scanlines is displayed, then 1/60 second later the field containing all the even scanlines is displayed. You never see both fields at the same time.

When video is captured by a computer consecutive pairs of fields are joined together to make full frames. If nothing moved during that 1/60 second interval you get a nice clean picture. But if anything moved you'll see interlace comb lines.

Where things are moving there's no way to perfectly create one whole picture out of the fields. Some deinterlacing techniques blur the two fields together (which works OK when motions are small, but when motions are large the frames look like double exposures). Some simply throw away one field and replace it with a copy of the remaining field or data interpolated from it (which effectively reduces the video from 60 pictures per second to 30 pictures per second and introduces some jaggy edges). Some programs try to figure out which parts of the picture contain movement and only deinterlace in those areas.