AVS Forum banner

1 - 20 of 24 Posts

·
Registered
Joined
·
442 Posts
Discussion Starter #1
Could use some help here from GraphEdit/FFDShow experts !


Have pretty big collection of SD DVD rips. Currently use FFDShow upconverting to play SD DVDs through MPC. While I am pretty satisfied with quality, I was always wondering if it is possible to make it even better.


CPU obviously puts certain limitations on real time processing, even my Conroe 6600/2Gig RAM limits scripts I can run (I upconvert to 1080 so it take lots of power).


So I am leaning towards doing 2 step processing:

1. Run CPU intensive FFDShow pre-processing (the best filter scripts) through GraphEdit and save original video stream to processed and improved video stream.

2. Upscale and run additional less CPU-intensive filter in real time when watching a movie.


Has anybody tried that before ? I searched AVS and other forums and have not come across anything similar.


My first attempt came to a halt with strange issue of being unable to write processed stream to a file -- when I am trying to connect Out pin of FFDShow Video Decoder to In pin of 'File Writer' I am getting an error that 'Filters cannot agree on a connection' (see attached image) -- Am I missing some filter that should be in between ? I saw lots of graphs where ffdshow audio is successfully streamed to 'File Writer' but I have never seen a graph of FFDShow video decoder working with 'File Writer'.


Need your help with FFDShow to 'File Writer' problem !


Also appreciate any advice about the 2 step processing idea, is it really possible to make SD DVD better with 2 step processing or am I wasting my time ? Thanks, Davie.
 

·
Registered
Joined
·
1,265 Posts
1. you're talking about re-encoding the video? avisynth scripting, ffdshow, virtualdub, or a mpeg2 encoder (hcenc), videohelp.com

2. ? just some sharpen, a little blur/nr ?
 

·
Registered
Joined
·
442 Posts
Discussion Starter #3

Quote:
Originally Posted by bulls /forum/post/13002415


1. you're talking about re-encoding the video? avisynth scripting, ffdshow, virtualdub, or a mpeg2 encoder (hcenc), videohelp.com

Yes, re-encoding and enhancing MPEG2 video by using FFDShow scripts. I thought the best way to do it is through GraphEdit, can it be done using other applications you mentioned ?

Quote:
Originally Posted by bulls /forum/post/13002415


2. ? just some sharpen, a little blur/nr ?

Second step is upscaling to 1080 and applying less CPU intensive scripts (incl sharpen)
 

·
Registered
Joined
·
377 Posts
FWIW:


for realtime playback, Dscaler5(+IVTC mod) + Reclock does wonders on 480i/1080i, can actually playback the 24p within. Not sure if it really helps with the upscaling part.


This, plus a 24|48|72|96|120hz LCD makes it look great.


p.
 

·
Registered
Joined
·
442 Posts
Discussion Starter #5

Quote:
Originally Posted by protovision /forum/post/13003171


FWIW:


for realtime playback, Dscaler5(+IVTC mod) + Reclock does wonders on 480i/1080i, can actually playback the 24p within. Not sure if it really helps with the upscaling part.


This, plus a 24|48|72|96|120hz LCD makes it look great.


p.

As I mentioned my goal is to make quality even better -- I am pretty happy with LSF and Seesaw scripts performance, but unable to run fft3dfilter script. What I am trying to do is pre-process MPEG2 stream with fft3dfilter (or any other CPU-intensive script) and then apply LSF and Seesaw on top of it. I think it should look gorgeous. With my current setup (Conroe 6600) I cannot possible combine 2 scripts for real time processing
 

·
Registered
Joined
·
442 Posts
Discussion Starter #7

Quote:
Originally Posted by Socio /forum/post/13003789


If you are going to filter and re-encode you might want to look through this old thread for some ideas.

http://archive2.avsforum.com/avs-vb/...hreadid=478263

Socio, that is exatly what I am trying to do
I already have all the tools (except DVD rebuilder) installed. DVD Rebuilder seems to be an interesting piece of software, it appears to go a step futher than GraphEdit, I could not find any filters that could mux audio and video back together so I knew I had to deal with it sooner or later. DVD Rebuilder seems to be an answer



Thanks for the thread and link ! Davie.


PS. Are you still rebuilding your DVDs or did you find some flaws with the process ?
 

·
Registered
Joined
·
2,403 Posts

Quote:
Originally Posted by eclipse98 /forum/post/13004235



PS. Are you still rebuilding your DVDs or did you find some flaws with the process ?

I stopped a while back, I am waiting until I can get a HD DVD, BD or Combo burner cheap enough. Then want to start doing it again on DVD's I know I will never see a HD version of, filter and up them to 1080P and burn them on HD DVD or BD for the next best thing.
 

·
Registered
Joined
·
442 Posts
Discussion Starter #9

Quote:
Originally Posted by Socio /forum/post/13004431


I stopped a while back, I am waiting until I can get a HD DVD, BD or Combo burner cheap enough. Then want to start doing it again on DVD's I know I will never see a HD version of, filter and up them to 1080P and burn them on HD DVD or BD for the next best thing.

I have HD (combo drive) on my HTPC, but still find myself watching a lot of SD DVD material (probably 70% or so) due to limited availability of movies in HD. Pre-processing DVDs makes lots of sense IMHO -- storage is very cheap these days
 

·
Registered
Joined
·
63 Posts
I convert my DVDs and over the air digital HD recordings to a common format. I re-encode the video to H264. I pull out the audio streams in their original format. Then I mux them into an MKV file, along with cover art, and information about the movie/program. MKV also lets me keep subtitles for DVDs that have foreign language tracks.


My display is 1080P native. So I have tried various things in the re-encoding to see what got the best scaled up picture in the end.


I found that doing significant scaling during re-encoding (upscaling all my video to 1080P encoding) may produce artifacts that get accentuated by the re-compression. Definitely leave final scaling for playback time.


I do however crop and manipulate the DVD scaling so that the encoding is done on "sqare pixels" i.e. not animorphic and even multiples of 16. For instance after cropping off the letterboxing, and scaling for proper 2.35 aspect (as close as I could with multiples of 16) I end up at a resolution of 848x360. I left the vertical dimentions of the pixels alone, then scaled the horizontal, so it was as close as I could get to 2.35*vertical. I use the Bicubic resize, with settings that don't sharpen of soften. I try leave the original video pretty much as is. But I have some DVDs where I probably will do some de-noise, and maybe some sharpening/smoothing.


A great advantage of pre-processing is figuring out what de-interlacing has to be done to get to a progressive video. I found that there are plenty of DVDs that will confuse the heck out of "automatic" or "smart" de-interlacing.


Since I use my HDTV display as my desktop as well as for movie playback, I tried converting videos from "video levels" to "PC levels" during encoding. Doing this has an advantage that if you are using VMR9 and it is using your desktop settings, you can calibrate your display and driver settings once and have everything display the same. I found however, that this does have slightly increased banding over having the playback software expand the levels at playback time.
 

·
Registered
Joined
·
442 Posts
Discussion Starter #11

Quote:
Originally Posted by cmonster32 /forum/post/13012561


I convert my DVDs and over the air digital HD recordings to a common format. I re-encode the video to H264. I pull out the audio streams in their original format. Then I mux them into an MKV file, along with cover art, and information about the movie/program. MKV also lets me keep subtitles for DVDs that have foreign language tracks.


My display is 1080P native. So I have tried various things in the re-encoding to see what got the best scaled up picture in the end.


I found that doing significant scaling during re-encoding (upscaling all my video to 1080P encoding) may produce artifacts that get accentuated by the re-compression. Definitely leave final scaling for playback time.


I do however crop and manipulate the DVD scaling so that the encoding is done on "sqare pixels" i.e. not animorphic and even multiples of 16. For instance after cropping off the letterboxing, and scaling for proper 2.35 aspect (as close as I could with multiples of 16) I end up at a resolution of 848x360. I left the vertical dimentions of the pixels alone, then scaled the horizontal, so it was as close as I could get to 2.35*vertical. I use the Bicubic resize, with settings that don't sharpen of soften. I try leave the original video pretty much as is. But I have some DVDs where I probably will do some de-noise, and maybe some sharpening/smoothing.


A great advantage of pre-processing is figuring out what de-interlacing has to be done to get to a progressive video. I found that there are plenty of DVDs that will confuse the heck out of "automatic" or "smart" de-interlacing.


Since I use my HDTV display as my desktop as well as for movie playback, I tried converting videos from "video levels" to "PC levels" during encoding. Doing this has an advantage that if you are using VMR9 and it is using your desktop settings, you can calibrate your display and driver settings once and have everything display the same. I found however, that this does have slightly increased banding over having the playback software expand the levels at playback time.


Very interesting info, thanks for sharing



So you re-encode your SD DVDs to H264 ? How does size of processed DVD compare to the original ? Is there any loss in quality ? Do you mux them into MKV because this is a container for H264 ? What tools are you using are you using for pre-processing ?


Lots of questions, appreciate your help
 

·
Registered
Joined
·
63 Posts
No problem glad to share.


Yes, I re-encode everything
There are a couple reasons. First, to get a clean and reliable playback file. With various DVDs, I have found that I "WILL" run into DVDs that give my normal DVD software issues playing them from the hard drive. A menu won't display, or a video will just stop halfway through the movie. So I convert to "MY" format and I can make sure it plays smoothly. Over the air broadcasts in HD are LARGE and sometimes have errors in what I got (might get a glitch in the image if you were watching it live). So I remove commercials, and the parts of other programs on the front and back of the recording. In the process of getting just the video/audio stream I want (I use VideoRedo) any issues in the video stream get cleaned up before I try to re-encode. The second reason is size. I want ALL my media available on my PC. That will require a large amount of disk space. But if I can use 15-80% less space than the orignal for almost original quality I see it as worthwile. With HD broadcasts, I use a fairly high bitrate (8Mb for 720P and 6Mb for 1080i), I don't save as much space on the 1080i as on the 720P, but the picture comes out looking pretty close to the original. The compression level on the 720P is actually higher, but the movies broadcast in 720 seem to suffer less than the nature documentaries that are in 1080i. With DVDs, I use two different rates.

One is for "live action" which is 4Mb. That is probably higher than I "need" but it seems to handle different content quite well. I took the "Lord of the Rings: Fellowship of the Ring" full extended version (almost 4 hours) from about 13 Gigs for the vobs, to a little over 9 Gigs, keeping the DTS soundtrack, and all 4 commentary tracks.

The other is for "anime" I would use this level for any traditional cell type animation i.e. cartoons. I use 1Mb for these. This is where I can take a DVD with 5 Gigs of video, and compress it to around 1 Gig of MKVs (with multiple audio tracks especialy 5.1, usually 1.2-1.3 Gigs).


Is there loss in quality? I would be outright lying if I said no. I am unencripting a lossy format, and re-encoding with another lossy format. There has to be a degredation. The question is how much, and is it perceptable in normal playback situations. If you want to do the right sort of test, you can find artifacts, and you can cause issues. Part of the reason I stated not to do full re-scaling before re-encoding, is I did screen captures from a test pattern and saw where there was some artifacting. I hadn't noticed that in the real conversions I had done, but decided to change my practice to avoid the chance of having it show up. Also encoding a DVD resolution video runs a lot faster than encoding a 1080P resolution video. Since scaling can be done very well at playback time, I wait till then to scale.


When I first ran into fansub anime in MKV it was a bit of a pain, because there were few players that could handle it. Now with the Haali splitter, and ffdshow (although I use CoreAVC for the H264 video since it uses less CPU) its fairly easy to get various players to play them (I am currently using Zoom player). I like the format for my own archiving, because these reasons:


Pack various types of video into it (although I only use H264)

Pack multiple audio streams in.

Pack multiple subtitles into it as well (useful for multilingual anime disks)

Pack attachments in - I am designing my own frontend, and want to keep cover art, and info about the movies all in the same file

Put in chapters if I want, this is nice with DVDs with multiple episodes from a series, you can create one MKV with all the episode, and using a chapter navigator in Zoom player or VLC I can jump to the episode I want



As for the tools I use:


AnyDVD - important for the DVDs with challenging protections schemes, regulary updated (I use this to either enable DVD Shrink to read the disk, or I use it to rip a full "clean" version to the drive, that I then use DVD Shrink on)


DVD Shrink - shame its no longer developed, but it does a good job of letting me pull out just the pieces I want off the DVD. I set it to "no compression" for the video, and I have it set with a custom DVD size of 15000, and have the option to split into 1 Gig vobs, disabled. That means that when I pull a large movie off a disk, the movie goes in one vob file, which makes the rest of the processing easier.


VideoReDo - useful for handling the HD broadcasts (mainly)


MeGUI - wonderful front end for a bunch of video/audio tools. I use its tools to strip the audio tracks out of my input file, and convert the video over to H264

(note, you need to install AviSynth seperatly, so that MeGUI can setup all the pieces you will want available)


mkvmerge GUI - part of the MKVtoolnix package. MeGUI will actually install a copy of mkvmerge and has a front end for it. If you just want to merge one or two audio tracks with a video track, that works quite well. Since I build rather more involved files (mainly the attachments) I use a seperate version and GUI for the merging.


I use other tools to deal with sbutitle stuff. Let me know if you want info about that as well. This post got a bit long
 

·
Registered
Joined
·
442 Posts
Discussion Starter #13

Quote:
Originally Posted by cmonster32 /forum/post/13015638


Is there loss in quality? I would be outright lying if I said no. I am unencripting a lossy format, and re-encoding with another lossy format. There has to be a degredation. The question is how much, and is it perceptable in normal playback situations. If you want to do the right sort of test, you can find artifacts, and you can cause issues. Part of the reason I stated not to do full re-scaling before re-encoding, is I did screen captures from a test pattern and saw where there was some artifacting. I hadn't noticed that in the real conversions I had done, but decided to change my practice to avoid the chance of having it show up. Also encoding a DVD resolution video runs a lot faster than encoding a 1080P resolution video. Since scaling can be done very well at playback time, I wait till then to scale.

Wow, that was long but insanely helpful, thanks for taking time to write it, I'm sure many people will find it very useful.



While your process perfectly fiits your needs (and I am sure I will use some of your tools/techniques), it seems our goals are quite different. I am not sure what display you use, but I project to 126" screen with 1080 projector (AE2000) hence even a small degradation of image quality is not acceptable. My goal is to improve image quality rather than decrease file size or add features, I would gladly accept 50-100% increase in media size if it provides me reasonable improvement in image quality.


I am curious as to why you would not try to apply some AVISynth filters to your SD DVD when converting them to H264 ?


Thanks again for your help !!
 

·
Registered
Joined
·
63 Posts
My display is currently a 37 inch Mitsubishi LCD TV. Using a DVI connection from the computer to the TV. I have two viewing locations. Right now I am sitting, about 6 feet away, so I can easily read web pages, and such. I also sit in this location to watch movies if my folks have come over to my place, and they are sitting on the couch. My couch (which is where I usually watch from), is about 10-12 feet away from the screen. Technically, sitting where my coffee table is, would be the ideal distance, but hey, I live in an apartment, I currently make do with the room the way it is.


Would some filtering improve image quality for my DVDs? It might. However, my thinking is that for the conversion from MPEG2 to H264, I want to do as little to the video as I can, so I don't cause more artifacting. I have read some posts about filtering, but I have never really sat down and experimented with it to see what I would or wouldn't like. I figure, if what comes out of the decode of the H264 is almost identical to what came out of the original MPEG2, then I can always post process, and get a similar look to if I had preprocessed in the conversion process. The advantage is, that if I decide I don't like the look of a particular post processing, I would just disable it, as apposed to trying to undo it once it was already applied during the conversion. As an example, after some testing, I found that the fullscaling with Lancoz and video level expansion I was doing was introducing artifacts. They only really showed up in the right sorts of video situation, but I didn't like it. There didn't seem a practical way to figure out what sort of processing I would have to run at playback time to eliminate them, so I am re-converting the videos I had previously processed (fortunitly only a small portion of my collection). Hence, I am doing as little to the videos as possible.


If you have a recommendation on some filtering that you think would improve quality, I would certainly take a look and try with a sample of videos to see what it does, and how it effects the final output. I certainly may see a dramatic difference and re-evaluate my stance on minimal manipulation during conversion.
 

·
Registered
Joined
·
442 Posts
Discussion Starter #16
I probably wouldn't bother with trying pre-process DVDs for 37" LCD, regular upconverting DVD player or HTPC with FFDShow looks almost like HDTV on my 42" LCD. When you go to 126" from 12 feet it is a completely different ball game. Socio posted and excellent link (on top of the thread) to archived thread that has a ton of good info how to improve video quality with AVISynth filters and DVD Rebuilder if you want to try. In your case I would at least give FFDshow a try (see sticky on top of this forum), I am using it right now with very good results
 

·
Registered
Joined
·
1,262 Posts
Here's how I would do it: Link to Instructions.


Just Make Sure that you Do NOt Resize in the Encode. It's not worth it. It will increase the file size to twice wjat it would be without it.


It takes a Very Long time to Encode too. 5 hours or so a Movie.

I suggest encoding your movie with fft3dfilter, then running the avi in mpc and using seesaw on top of it in realtime. If you want to use both. Otherwise the encode might take TEN Hours or MORE to do it all at once. This is with a E6600 CPU too.
 

·
Registered
Joined
·
442 Posts
Discussion Starter #18

Quote:
Originally Posted by 8:13 /forum/post/13032223


Here's how I would do it: Link to Instructions.


Just Make Sure that you Do NOt Resize in the Encode. It's not worth it. It will increase the file size to twice wjat it would be without it.


It takes a Very Long time to Encode too. 5 hours or so a Movie.

I suggest encoding your movie with fft3dfilter, then running the avi in mpc and using seesaw on top of it in realtime. If you want to use both. Otherwise the encode might take TEN Hours or MORE to do it all at once. This is with a E6600 CPU too.

Thanks 8:13, I will give it a try. Since you are our local FFDShow expert, I would appreciate your feedback on how much picture quality improvements can I expect by pre-processing with fft3dfilter and then playing real-time with LSF and upscaling ? Is improvement going to be noticable and worth spending time and effort ?


Also, I am curious if you tried using your method in DVd Rebuilder, your AVISynth script looks like it might work well with DVD Rebuilder, so why going all these extra steps (installing apps, etc) while one can simply plug it in in DVD-RB and it will do the rest ?


Thanks for your help, Davie.
 
1 - 20 of 24 Posts
Top