Help encoding mp4 with lots of action (wrestling)

[owen316]

I am looking to convert my collection of wrestling dvds to mp4. I have tried a number of programs but the results just aren't that great. There are usually a number of artifacts on the wrestlers and/or they aren't that sharp.

I have tried DVDFAB and Staxrip and they produce similar results. I use the high quality options, 2 pass encoding and the file size is around 700mb for 40 mins or so.

Does anyone had any success with this? Or have any ideas?

[poisondeathray]

The dvd's probably look like that to begin with . mpeg2 will likely be full of artifacts, macroblocking in action scenes

You might need to preprocess with some deblocking filters

Post a screenshot or sample , or describe the "artifacts" more descriptively

[owen316]

The DVDs are actually pretty clean. A lot are converted from master VHS all in SP. However can you please explain "preprocess with deblocking filters?"

[poisondeathray]

Filters are applied before the encoder, so the encoder can do a better job with compression . Noise, macroblocks, these are all detrimental for compression . MPEG2 compression (used in DVD) isn't very good

Depending on how you did the VHS transfer, I suspect you can benefit from filtering it. Even a retail digital (non analog vhs) wrestling DVD will have artifacts and can benefit, so a VHS case would likely benefit even more

It's either that, or you need higher bitrates if you want to preserve the entire signal (even the noise)

[owen316]

Here are links to the files I'm working with
http://www.megaupload.com/?d=6PE03HS9

[poisondeathray]

link appears to be down

"The file you are trying to access is temporarily unavailable. Please try again later."

[owen316]

I edited the link. Sorry about that.

[poisondeathray]

you can upload it directly to the site if <30MB , or use another free file host e.g. mediafire.com, sendspace.com

just need a short representative sample of the source that you're dealing with

[owen316]

Its just about 50mb. Its a 1 min clip of the vob and the mp4

[poisondeathray]

The DVDs are actually pretty clean. A lot are converted from master VHS all in SP. However can you please explain "preprocess with deblocking filters?"

It's not that "clean" at all. It's your "typical" VHS transfer, with typical quality issues. There's quite a bit you can do in terms of preprocessing. A better capture (if possible) is ideal, because this transfer is low quality. There is significant macroblocking and pixellation , noise. All types of noise are taxing for encoders , as they attempt to preserve that noise. As a result you get compression artifacts, lower quality. You need higher bitrates for similar levels of quality. That's why people preprocess with denoising filters.

This can be a pretty big topic, so I'll try to be brief and introduce some "key words" that you can use to search for more information and/or guides

Your basic problems:

0) The aspect ratio of your encoded file is off; it's 3:2 ,but it should be 4:3

1) Not an ideal capture - there are macroblocking issues (mpeg2 at DVD bitrates) . A lossless capture wouldn't have those issues. You can use deblocking filters, but they all blur the image to an extent

2) Noise, chroma noise, luma noise - these require high bitrates to encode, that's why people pre-process with denoising filters. Chroma noise - if areas that are black or uniform color, have "rainbow" colors . Typical for VCR's transfers , you can use chroma noise filters

3) Deinterlacing quality . You're single rate deinterlacing to 29.97 fps (throwing 1/2 the information out). The quality of deinterlacing varies tremendously. There is a quality/speed tradeoff some are very fast but lower quality like the one you used (yadif). This is a huge topic, you can use search, as interlace / deinterlacing is discussed thoroughly in other threads. Avisynth has a large selection of deinterlacers, and the best one is probably QTGMC

4) Black level is too high => gives a "washed out" , low contrast look . Luminance Y' should be between 16-235 . You can use "levels" filters or smoothlevels in avisynth

5) Halos - some parts have thick outlines, there is not much that will take care of this . There are dehalo filters for the thin outlines

6) Persistent lower right line defect - you might be able to use inpainting , mask this out or use delogo filters




Here is a filtered sample, it has 2x the number of frames, 2x the fps, yet uses the same video bitrate 1.8 Mbit/s. This is possible because it was denoised. Encoders generally encode the differences between frames, so smoother video requires less bitrate. Things like vast changes, noise, require higher bitrates

I used avisynth filters for this, here is the script. As usually adjust the settings to your own tastes

MPEG2Source("VTS_01_1.d2v", cpu=3)
AssumeTFF()
QTGMC(preset="faster",sharpness=0.5)
SmoothLevels(25,1.2,255,0,255)
MCTemporalDenoise(settings="low")
FFT3DFilter(sigma=3,plane=3,bt=2,ow=16,oh=16)
TTempSmooth(maxr=5, lthresh=8, cthresh=4, strength=5)

*A lot of this is subjective - some people prefer very clean images, calm and stable - but the tradeoff is less fine detail. Some people prefer the original noise. There are tradeoffs for everything, in terms of processing time, quality. A "cleaner image" means less detail, but subjectively might be more appealing to watch

Also, instead of using a fixed bitrate, consider using CRF or quality based encoding. You get the same average quality and the encoder will adjust the bitrate appropriately. So action content will get higher bitrates. A slow moving interview piece would get lower bitrate

[owen316]

Thank you! I think that sample looks great for the DVD.

I think I am lost on how to use avisynth. Also after it has been run through avisynth does it matter what encoder I use? Could you give me a quick step by step on how you got the final product?

[jagabo]

The top class software for filtering and encoding is all free.

[poisondeathray]

I think I am lost on how to use avisynth. Also after it has been run through avisynth does it matter what encoder I use? Could you give me a quick step by step on how you got the final product?

There is a bit of a learning curve, start with the basics:

http://avisynth.org/mediawiki/Main_Page
http://avisynth.org/mediawiki/Main_Page#New_to_AviSynth_-_start_here

You have to gather the plugins, prerequiste dll's and avs scripts before you start

All the filters are named, so if you search on google , you can find them. Most of them can be found on Doom9 (or links to where the filters are hosted). For example, QTGMC has many dependencies, and the Doom9 thread has a link to the various plugins in the 1st post
http://forum.doom9.org/showthread.php?t=156028

Most filters have included documentation & instructions



I prefer applying the filters and encoding to a lossless intermediate , then using that intermediate as input into the final encoder. I used x264 in that example, there are many GUI's for it e.g. megui, staxrip, xvid4psp, ripbot, handbrake etc...
I prefer this way for 2pass encoding, because slow filters are applied once instead of twice . When using slow filters, the overall process is faster. If you are doing 1pass CRF encoding, then it might not make sense


There are many ways to get similar results, you might use different filters, or even non avisynth filters. Vdub has a bunch of filters that are similar. You might use something like neat video for the denoising

[jagabo]

I usually use a batch file to encode with x264. I have one in my SendTo folder so I can right click on an AVS script and select SendTo -> (appropriate batch file).

Start /b /low x264.exe --preset=veryfast --crf 18 --ref 3 --bframes 2 --keyint 100 --sar 32:27 --output %1.mkv %1
pause

That gets you about 90 percent of the compression that x264 can deliver (at CRF 18) with very good quality. Yes, I could get a 10 percent smaller file by using slower settings but why bother.