1080i downsize filter?

[hollywood10s]

I have been encoding some videos for my website using Megui and my original file size is 1080i (1440x1080) I have been downsizing them to 720x405 for playback on my website here is an example (http://www.steelheadstalkers.com/egg_loop_knot.htm). The video in fullscreen looks pretty good as well (this videos is encoded at 500kbps) Now my question is should I only downsize to 1280x720 to make better quality videos and what filter should I use to downsize my videos (lanczos4, spline36, ext.....)?

Will I get better percieved quality only downsizing to 1280x720 instead of 720x405? I have been using lanczos4 to downsize and Im just wondering if there is a better way. Also if I downsize to 1280x720 instead of 720x405 will the video look better or worse at the same bitrate (500-1000kbps)? Some of my videos are still (not much motion) others are of fishing (more motion and moving water) so maybe I need to use different techniques for both. Thanks for all of the help.

Chad

[Blue_MiSfit]

If you're going to downscale 1080i to 720p, don't forget to deinterlace first!

Should you go to 720p? If you like the look of it better, and can afford the extra data rate, and aren't worried about your viewers' PCs having the horsepower to play it.

If you're going to do it, you need to try different resizers, we can't tell you which one will be better or worse :) I would say Lanczos4 might be a poor choice, since it tends to be a little too sharp for big downscaling - but again that's my opinion. Spline16 / Spline36 are very neutral. Lanczos might be a good bet. Try!

If you go to 720p versus 405p at the same bitrate, you will probably see more artifacts. 500-1000kbps is a little low for high motion content at 720p, no matter what codec you use.

Lots of work will answer your questions :D

~MiSfit

[hollywood10s]

Thanks for the reply. Until my quad core comes in and I get ot all set up. Testing takes a lot of time, so I thought I would try to get as much good information and read up before I start the hour long tests on ten second clips :)

Another question. Would I get better picture quality by downsizing 1080i (1440x180 de-interlaced and downsized to 720x405) or downsizing 720p (1280x720p downsized to 720x405)? I figure the second might produce better quality because it is being downsized less going from 1280x720 to 720x405 instead of going from 1440x1080 to 720x405. Let me know what you guys think. I can export my raw files to either 1080i or 720p. Thanks again

[audyovydeo]

Until my quad core comes in


Let me add my 2€cents to Blue_misfit's : people making video for any form of distribution (like me) should buy pentium4, not quad-cores. It's the only way to make sure it plays back on everyone's hardware ;-)

cheers
a/v

[Blue_MiSfit]

Well, you should use a quad core to do your work on, and a cheap P4 / Athlon64 / Mac to test your stuff! No reason to gimp your workflow.

Now, as to your scaling question. It depends on your raw files. What are they?

[hollywood10s]

They are 1440x1080 AVCHD files. Let me know if you need any more info on the file type. I have been editing them in ulead video studio 11 and them exporting them to either 1440x1080 HDV or 1280x720 HDV to import into Megui.

[Blue_MiSfit]

Here's a suggestion. Try to output to a lossless file if you can. AVCHD is MPEG-4 AVC, and when you edit and output to HDV, you're recompressing to MPEG-2. Then in AviSynth / MeGUI you're filtering / resizing etc, and re-encoding _again_ to (guessing) MPEG-4 AVC (H.264) again for delivery.

Try to minimize the number of compression steps.

The wonderful Debugmode frameserver for Vegas / Premiere doesn't work on Ulead sadly - but if Ulead has a half decent VFW export module you should be able to export your project to a Lagarith AVI (at 1440x1080), then load that into AviSynth and do all your processing.

I wouldn't trust whatever deinterlacing / scaling routines ulead uses. Even the big dog commercial NLEs like Avid and Final Cut have piss-poor deinterlacers by default. Magic Bullet and the like are pretty good, but AviSynth is always better! :) If it's in your workflow, you might as well maximize its potential.

~MiSfit

[hollywood10s]

I have been doing all of my deinterlacing through Megui using avisynth. Lagarith AVI I believe I have tried. Are the file sized very large outputting to Lagarith AVI? It doesnt matter to me how big the file is if Im going to get a better final product. I will try extracting my files to Lagarith AVI then load that file into avisynth through megui. Lossless means 100% quality, am I correct? So the least amount of times I have to compress, the better final quality I will probably have? Thanks for all of your help guys

[thetoof]

Yeah, in this case it's lossless compression (like a .zip or a .7z), meaning that regardless of the times you compress it with Lagarith, you won't lose any quality.

[Blue_MiSfit]

Yes, Lagarith is very large. Mathematically lossless means large file sizes. It typically averages 450kb per frame for 720x480 video for me, which for 24p means about 86 megabits per second. It will be a lot more for 1440x1080.

Basically, the uncompressed video that Ulead renders after your edits etc go straight into Lagarith, which compresses them. The output into AviSynth from Lagarith is _exactly_ what was output from Ulead. The only disadvantage is the large intermediate file.

The perfect solution is to have a frameserver - which just outputs uncompressed video straight from Ulead to AviSynth, but sadly there doesn't appear to be one that works with Ulead.

All your other guesses are absolutely correct.

Make sure to output 1440x1080i from Ulead, as I mentioned. Try to let Ulead process the video as _little_ as possible, since you don't have explicit control over that processing. With AviSynth you do - that's the beauty of it.

~MiSfit

[hollywood10s]

I also have vegas 8.0b. I havent spent much time using it but I will eventually. Using vegas are there better ways of achieving what you are talking about? Sould I load all of my AVCHD files into vegas and output them to avi? Also should I choose uncompressed for the video format in AVI using vegas? Thanks

[Blue_MiSfit]

Do not use uncompressed AVI for the video format in Vegas if you're going to use that.

Leave it as is, but install Debugmode Frameserver (google). When you select this codec for output, Vegas will create a "dummy" AVI that can be loaded into AviSynth. Filter as necessary and encode :)

~MiSfit

[2Bdecided]

I would check a file of the required resolution, bitrate, frame rate, and codec on that Pentium4 to see how it plays before spending days optimising a workflow to generate the "perfect" file.

You will probably find you are asking too much of many people's PCs. If your users are used to 720x405, 1280x720 is going to come as a shock to some with older PCs.

For example, I can play HDV (1440x1080i50 MPEG-2) fine on my P4 2.8 machine, but not 1280x720p25 with some newer codecs, and certainly not 1280x720p50. I can't even play HDV smoothly on my P4 2.0 machine without dropping frames. 1024x576p50 MPEG-1 or MPEG-2 works fine though.

As an example of furious users who can no longer view video smoothly, you should have seen what happened when vimeo allowed 30fps instead of 24fps. It pushed so many people's machines over the edge that they had to relent, and go back to re-encoding 30fps to 24fps so that it played "properly".

Cheers,
David.

[hollywood10s]

You can check an example here to see if your computer can play the files I am creating. (http://www.steelheadstalkers.com/float-fishing_tutorial.htm) By the way I'm testing this website on a AMD Athalon 64 3500. I have been encoding at 720x405 and when played in fullscreen it is upscalled to 1280x720 (I think) with marginal quality loss. These files should play fine on pentium4 and I would like to hear if you have a problem with the example above. Everyone that has viewed the videos that I have heard from has not had a problem. I am just trying to squeeze a little more quality out of my files and if I can do this by adjusting my workflow then I will.

One more thing. Once frameserver starts I have to keep it running in the background correct? So I have lets say 15 clips in AVCHD that need to be edited and then put together in vegas or ulead to make one video file. Should I edit and put them together as AVCHD in vegas or ulead and then use frameserver to make the final file in Megui/avisynth? Can I do the same with files that are already encoded to HDV? Export them to frameserver (lossless avi)? Maybe I am doing something wrong, should frameserver save the file to my computer for me to encode later or does the program (frameserver) have to be running in the background? Thanks for all of the help.

[2Bdecided]

That plays fine here, but that's the resolution that you've been using all along, isn't it? Apart from the jaggies when scaled to fullscreen, the quality is excellent. If only most web based video was even half this good! Mind you, you have an easy subject to encode.

FWIW I think it was crying out to be shot in 4x3, not 16x9, but that's just me!

Cheers,
David.

[Blue_MiSfit]

Yeah, that looks great to me!

I agree it would be better cropped down to 4x3, and it would be very nice if Flash could dither on playback (add some noise to hide banding)

Do the editing first, then frameserve your output into AviSynth. Keep the frameserver and editing program open while you encode, then when you finish you can close it.

~MiSfit

[smok3]

huh. nothing cryes out to be shot at 4:3...

[Mug Funky]

..except everything in 4 perf super35...

[hollywood10s]

Its original size is 1440x1080i. But it is normally upsized to 1920x1080 I guess, so this is why I put it in the container at 720x405. If I encode it in 4:3 will it using elongated pixels stretch it out to 16:9? Another thing I am trying to make better is the videos while fishing. I have to the bitrate up to 2000kbps because of the moving water and other movement (I believe it is harder to encode http://www.steelheadstalkers.com/applegate_video.htm). Is there any avisynth filters or other settings to allow me to go down to say 1500kbps or 1000kbps? What does (4 perf super35) stand for? The still videos like the example above are encoded at 500kbps and Im pretty happy with them.

One more question. When I output to debugmode avi is it saved a progressive file type? Or is it interlaced like the original? Thanks again for all the help.

[Blue_MiSfit]

debugmode outputs the exact video stream, whatever it may be. So - since you have an interlaced source it will output that.

4perf super35 is a film format :) Mug Funky was being facetious (I think)!

We just meant that your sample video wasn't really using the "Wide" space offered by the 16x9 aspect ratio. That scene would have been just fine in 4x3 :)

As far as filters go, there is no one filter that will do amazing things. Let's continue denoiser discussion in your other thread?

~MiSfit

[2Bdecided]

huh. nothing cryes out to be shot at 4:3...IMAX? ;)

Seriously, if you're going to watch on a widescreen TV, then you have little option.

However, if you're going to embed into a web page, you can use whatever you like. Even full screen on a PC, few people today have 16:9 monitors (most of those few that think they do actually have 16:10 - many who think they have 4:3 really have 5:4!).

Anyone who thinks that, in a hypothetical world of variable aspect ratio displays, everything should be shot in 16x9, clearly has no eye for composition, and has never visited an art gallery!

I won't mention how many of my 16x9 home movies end up being 4x3 safe automatically because of the fact that any natural scenes fit this area nicely! Not all, but most. "planned" content (films, TV, etc) is different, but... Oh dear, I said I wouldn't mention it!

Cheers,
David.

[smok3]

clearly you were annoyed by the central composition of shots, but that has little to do with 4x3 or 16x9, however 4:3 is an ugly ratio, almost anything is better than that, 1x1 for example.

and has never visited an art gallery!

that was a bit low wasn't it? (ok, ill go back to cow milking now...)

p.s. oh, i think most of the stuff could be shot at 21x9 actually :P

[2Bdecided]

Sorry smok3 - I wasn't blaming you for the 16x9 take over of the world. I think it's partly there to sell more TVs.

For example, 14x9 is also a versatile ratio, and one previously chosen for HDTV - now rejected - maybe it's not different enough from 4x3 for normal "dumb" people to realise they need a new TV? ;)

Anyway, back on topic - yes, the composition of the frame wasn't great for 16x9, though I doubt I'd have done better.

FWIW I've tried 1x1 for some web video on facebook, because the content looked like it would work, but it really didn't. That is ugly - on that particular page, 4x3 looks "square" so 1x1 looks bizarrely more tall than wide.

I guess for web video, just for efficiency, it makes sense to crop out unused parts of the frame; but if nothing moves in those parts they don't take many bits with long GOP formats anyway.

Cheers,
David.

[smok3]

actually some broadcasters seems to use 14:9 as broadcast only format (since it is center between 12:9 (4:3) and 16:9), i did various tests in the past comparing 16:9LB and 14:9LB and for my eyes (even on small 4:3 display) 16:9LB always wins..., even worse, i thought 14:9 compromise is silly, but generally that is all debatable.

edit: what's wrong with my natural looking avatar at 1:1? I guess motion has something do with perception of aspect ratios?

edit2: history: wide formats were adopted by cinema to take over the crown from those days tv's, now the tv will punch back (finnaly).