Encoding for the "big screen"?

[surreal120]

Just put together an HTPC setup and want to play back by encoded videos (DVD -> Xvid or x264) on my TV, which is a 720p LCD.

Should I then be resizing my videos to *larger* than anamorphic in my AviSynth scripts (i.e., should/can I encode directly to 1280x720)? Also, does it still make sense to use Lanczos for playing back at such a large size? Or would something like Bicubic, or even Bilenear give better results?

[DarkNite]

I encode to 720p all the time. If I'm using SD sources for those encodes I upsize using LimitedSharpen (there's also LimitedSharpenFaster) or iiP depending on the condition of the source. LanczosResize would certainly be faster, but LimitedSharpenFaster isn't going to bog you down so much that the results aren't worth the time involved. iiP is a different beast entirely, and not for the faint of heart, but can definitely deliver quality results. Try a few test sections to compare methods and see what works best for you.

[surreal120]

I've never used LimitedSharpen before for resizing. Would you mind posing a sample of your avisynth script?

(Also, and I know this is a *REALLY* dumb question, but what's the easiest way to limit encodes to a sample of the video source? Encoding time hasn't been a huge priority for me, as I'm still just learning the basics of how to make a quality encode, so I've pretty much just been using MeGUI with the hq-slowest profile and haven't experimented very much with avisynth yet).

[Awatef]

Just wondering if you really have a 1280x720 LCD. I see many posts here of people wanting to resize SD sources to 1280x720, but I have yet to see an LCD on the market with this resolution. Most manufacturers tend to use 1366x768 screens...

Here again, I want to remind people that reencoding a DVD to make it HD-like will make you LOSE details, not gain some. There is always the way through ffdshow to use better resizing types like Lanczos, instead of bilinear or bicubic.

[surreal120]

Just wondering if you really have a 1280x720 LCD. I see many posts here of people wanting to resize SD sources to 1280x720, but I have yet to see an LCD on the market with this resolution. Most manufacturers tend to use 1366x768 screens...

My TV has a DVI input, so it can function as a monitor at 1280x720. I can also hook up my video card to my TV via component connectors, in which case the display size is actually slightly smaller than 1280x720 due to overscanning. (In any case, I would gladly take a TV that is advertised as only 720p but actually has a *higher* resolution, e.g., 1366x768!)

All that aside, you may be right that reencoding at a higher resolution will lose detail in the end, if bitrate is held constant.

In that case, what do you recommend for encoding SD content for 1280x720 tv output? I've tried resizing in ffdshow, but I could not tell any difference between usng ffdshow's resizing and simply letting ZoomPlayer resize to fill the screen. Moreover, upsizing "on the fly" surely can't be as efficient as upsizing during the encoding process. No?

[surreal120]

I've been messing around with avisynth filters for quite some time now, and I'm feeling (a) overwhelmed and (b) like I know less than I did when I started! It's now 4:30 am, and I'm tired, frustrated, and confused.

If anyone wise sage out there once had that feeling at some point, hopefully you can help:

I'm trying to create high-quality x264 encodes of my DVDs for watching on an HTPC (which is connected to a 720p LCD TV). My source material is basically recent, progressive, NTSC DVDs. Every time I think I've got the hang of something in avisynth, I try changing one setting and suddenly nothing makes sense again...

For encoding newish DVD material, is a basic crop and resize all you need from avisynth? Or are there some "necessary"/"magic" avisynth scripts that should generally be used?

For the most part, the biggest issues I notice with my encodes seem to be color "banding" (that's how I would describe it, at least, but I may be using that term wrong - some areas of color don't fade smoothely into the surrounding areas) and *also* blockiness/"pixeliness" around the edges of faces, or shadowy areas of skin (like the area below a person's neck).

Is there an avisynth filter designed for either of these problems? Every time I try to search the forums and think I've found something, I usually just spend hours trying different settings, and eventually everything starts to look kind of the same. I'm lost and confused...

Please help!

[Awatef]

See? you got some "nice" artefacts there. :)
Color banding and/or blocks and/or mosquito noise and blurring are the price to pay if you want compact files.
If you want to minimize this (and I emphasize *minimize*, you can't avoid them completely!), you'll have to encode at a very high bitrate, which may yield to sizes beyond a normal DVD-R, so...

By the way, if your TV doesn't have exactly 1280x720, your TV-chip will kick in and resize again! so you'll have 2 resizers instead of one, which means 2 times quality "disturbance".
Besides, if you force the resolution on 1280x720, although your LCD is designed for another resolution, you'll get crappy quality, since LCDs work best only with their native resolution (I have one in front of me right now, so I know what I'm talking about :))

One last thing, if you don't see any difference between bilinear and lanczos resizing (which is no different from offline resizing btw), you can call yourself lucky, at least you can enjoy watching TV without worrying about missing something ;)

[surreal120]

I was so tired I forgot that I had ffdshow resizing with bilinear instead of lanczos! The resize filters are actually the one avisynth function that does make a huge visual impact for me (although my experience is limited). The reason I switched ffdshow to bilinear was that Lanczos was creating way too many artifacts - probably because, as you say, the TV is doing some resizing of its own. I always encode with lanczos, though (at SD resolution).

Another thing worth mentioning is that just playing the actual DVD exhibits many of the same artifacts as well. It's usually *slightly* better in terms of quality, but it's not noticeable in most scenes (obviously, darker scenes reflect the widest disparity between the encoded video and the source).

However, it seems as though many people in these forums talk about "cleaning up" the video in Avisynth. I had always thought this meant you could compensate for some of the artificating with avisynth if there are very specific issues. I don't know why I assumed that (I thnk I just made an "ass" out of "me")... If my encode looks pretty close to my source, is that about the best I should be hoping for? Something just seems "wrong" to me if all I'm doing in avisynth is crop and resize (and maybe IVTC if necessary), given the abundance of avisynth filters out there.

[DarkNite]

You can definitely clean up badly mastered DVD's, and you can improve upon even well mastered DVD's, but what looks good to your human eye often looks like a nightmare to a lossy codec.

Wish I had more time to help right now. There are a lot of good threads on this topic.