Thank you for posting a better info source. From the mediainfo report and from your earlier comments, a few things are possible. Some things that are not possible, however, are that this video is neither "HD" nor 16:9. It's not BluRay, it's not AVCHD. It's plain-vanilla standard-definition 4:3 DVD in a standard 720x480 DVD frame. The standards for DVD format require that in order to meet playback spec in most DVD playback devices, DVD should be encoded as either interlaced or as progressive+pulldown (aka "fake interlace"). Of course, any idiot with some video software can encode a DVD however they wish, along with screwing up processing on the way to encoding. Just because someone owns a copy of Premiere Pro or Vegas Pro doesn't mean that they know what they're doing.
Depending on how the data flags in that DVD are written (or not written, as the case may be), you could be looking at: a) interlaced or telecined video that has been encoded with progressive data flags. b) progressive + pulldown (telecined) video either encoded as progressive or simply reported that way by analysis software (the latter often happens because of the way playback flags are encoded in the video). c) Interlaced video that has been field-blend deintelaced. This is probably not the case, or the video would never appear to play back properly using forced deinterlacing in a media player. The main thing you should be aware of is that progressive video, whether originally recorded as progressive or properly deinterlaced into true progressive format, does not display "interlace artifacts". By interlace artifacts we assume you mean "combing" effects, especially during motion. It's possible for MediaInfo and similar software such as GSpot to report any of the above as Progressive. Since you don't appear to be expert in using software like Avisynth or VirtualDub to analyze and break down a piece of video, the only way to tell for sure is to submit a sample for inspection. You can use the free DGIndex, an easy to use and setup product developed for the purpose of reading VOB or MPEG files and saving them as computer video files or for cutting samples for analysis. It can extract a short sample without altering the original signal other than demuxing (separating) video and audio into separate files. The video portion is all we would need. About 10 or 12 seconds of demuxed mpeg (it would be a ".m2v" file) containing lateral motion such as someone or something moving from side to side, would be less than 32MB. DGindex is part of the free DGMPGDec package, available here: http://www.videohelp.com/tools/DGMPGDec . It's a free-standing program with no installer. Create a folder for it (Call it DGIndex) and unzip the download package into that folder. Inside that folder you'll see DGindex.exe, which is the extract program. Double-click that .exe to run it. A quick and simple tutorial on how to make a video sample is here: http://forum.videohelp.com/threads/3...=1#post2272359. The final upload steps in that tutorial are for the other forum, but the same procedure is identical here at digitalfaq -- except that you have to click "Go Advanced" at the bottom-right of our Reply window to see the "Manage Attachment" upload icon. |
How big should the file be? When I try to upload all I get is sending request to digitalfaq. Is there a size limit?
I'm not usually so thick when it comes to computers but I have had a look at the upload instructions and I can't get it to work. No matter what setting I enter in WinRAR it splits it all into 1kb files!! Is there any other way to get the file to you? I have a website. I can upload it there and you can browse to it I suppose. I got it to work with HJ Split. I don't know what is wrong with WinRAR. I'll upload in a bit. Well I still can't get it to upload--just sits there sending a request to digitalfaq. Let me try something else. I'm uploading the files to my website. I will post a link when that is done. Maybe it's an issue with your server. |
Quote:
I've been using Photoshop CS3 for ages now, along with Premiere CS4. I fired up Photoshop CS4 a few days ago, to take care of something quickly on the video system with Premiere CS4, instead of using my normal photo system with CS3. I found a bug in CS4 that prevented selection options, not present in either CS3 or CS5 (which is also here on yet another system). So sometime you can know what you're doing, yet the program does what it's not supposed to. I've seen this in Premiere as well, but it's been too long ago to give specific examples. Food for thought. :wink2: |
Here's the link where you can download the files. Just save link as... and all should be well. There are six altogether. I tested rejoining them with HJ Split and everything worked fine.
http://www.superlather.com/DGIndex files.htm I'd appreciate any help you can offer. |
1 Attachment(s)
Thank you for the samples. Thanksgiving slowed things around here, but I've had a chance to give them a good look this morning. Will have some info and new sampoles for myoiu soon in a short while.
BTW, the reason your samples wouldn't upload is because the server doesn't recognize file endings like ".001", ".002", etc. In any case you needed only several seconds of video but uploaded over 3 minutes. Anyway, I'm preparing a reply post now. Sorry for the delay. EDIT: Uh-oh, family duties interrupting things for a couple of hours. Back later today. About 3-PM now. -- merged -- Quote:
What HD camera made the original? I ask because the ISO extracted sample is 4:3 SD-DVD. HD cameras don't shoot 4:3 in hi-def frame sizes unless they're set for 1440x1080 or standard definition 720x480. Those are the only two HD formats that allow 4:3 DAR. All the other frame sizes are 16:9 only. Be that as it may, what you have is a case of either field-blending or frame-blending (looks like field-blend deinterlace to me, or it was resized while still interlaced, which results in the same artifacts). I have attached an edited deinterlaced sample below (vts_deinterlace_5fps.mpg). These are brief sequences of hand movements cut from your m2v sample. You commented that you saw satisfactory results when deinterlacing with VLC, but in fact no media player can properly deinterlace this video because it's not interlaced. The frame rate of the mpg attached below is slowed to 5fps to show what deinterlacing players are really playing. Most of the blending is in the chroma channels, likely due to the way YUV data is stored and interpolated when a video is processed in an RGB colorspace and re-converted to YUV improperly. What deinterlacing does is realign the "fields" (actually, the odd and even lines), which can often smooth sawtooth edges, while QTGMC helps to reduce combing artifacts. What you still see are ghost images from the blending, regardless of how the video is played. Another thing you'll note in the 5fps sample and in the original m2v is the amount of block noise, line twitter, and color corruption in the speaker's suit, usually a result of improper re-encoding. In all playback modes you'll also see a brief amount of motion stutter now and then. In Avisynth I used two runs of the Santiag anti-alias filter, along with QTGMC in progressive repair mode, followed by motion-compensation cleanup and motion smoothing (MvDegrain2), then a dehalo and edge bleed filter (aWarpSharp2), and a 16-bit dithered antibanding filter and conversion to RGB32/Rec601 for color correction in VirtualDub before setting the colorspace back to YV12 for encoding. In the 3-minute reworked sample at this mediafire link (https://www.mediafire.com/?s9n2dkxa3gnoqy1), 147MB), you still see some motion smearing in the guy's suit (which I guess is better than the blocky artifacts of the original), and some of the same occasional motion stutter that appears in the original -- it's just easier to see in the link because a lot of noise, sawtooth edges, and combing were cleaned. Whoever processed your original video is either very careless or just clueless. IMO it looks like a case of the latter. At least they might have fixed the color temperature, which isn't rocket science. |
First sanlyn I wanted to thank you for taking the time to do all of that. I hope you had a Happy Thanksgiving.
Quote:
Thanks again. |
sanlyn:
- Can you post that script? - And next time, can you please post the sample clips to the cdn4 ftp? You should have access. If not, ask me for new credentials. We don't want to lose them when the file disappears from the 'file sharing' site. It's a good 'after' clip, and you probably went further than what I would do. :) |
1 Attachment(s)
Quote:
I ran the script in two steps (well, 3 if you count encoding). It's neither a fast script nor especially slow, it's just easier to tweak filter values if some steps run separately. Avisynth and VirtualDub took about 35 minutes, and about 1.5 minutes for TMPGenc to encode. I use other encoders half the time, but TVMW5 sets up faster and gets pretty decent results. I used XP SP3 with an Intel i5-3570. I actually spent more time than that. The script was adapted from an older one. I spent about 2 hours trying other ideas with Avisynth, some deblenders, and VirtualDub. Playing Dr. Frankenstein is one way to find out what some of these filters do. With more motion or camera pans, forget it. You can't do much with the permanent damage this kind of blending inflicts. For a really long video I'd run it while doing other chores, or just let it simmer overnight. Avisynth MT can often run faster, but not with all filters. Sometimes you spend more time fixing MT settings than you do running a script the "old" way. Quote:
BTW, I never did find my way to the ftp instructions. I'll try another look and will PM you (?) if my old brain gets lost looking for it. The 'after' clip is probably too long anyway. Quote:
A .vcf file for the VirtuaLDub filter settings is attached. Avisynth plugins: # - QTGMC-3.32.avsi and support plugins # - Santiag.avsi (requires NNEDI3.dll, which comes with QTGMC files) # - vInverse.dll (anti-combing) # - aWarpSharp.dll (v. 2012) # - Grain Factory3.dll # - AddGrainC.dll (comes with QTGMC support files) # - GradFun3 (function in dither plugins, see below) # - Dither_convert_yuv_to_rgb (function in dither plugins, see below) # ------------- DITHER plugins --------------------- # These are kept in a separate plugin folder to prevent # conflicts with other dll's having similar functions. # - Dither.avs # - mt_xxpand_multi.avs # - avstp.dll # - dither.dll # - mt_masktools-26.dll # ------------- MVdegrain_2B routine --------------- # a denosing routine posted in Doom9 and Videohelp by 2BDecided. # It uses older "v1" versions of masktools and MvTools. These # can be loaded explicitly with newer versions of MVtools and # MaskTools at the same time because the two versions have # different internal function names. Older MVdegrain often # considered "better" at retaining detail while denoising. # ->> The coded routine is in the "Step2" script. # - mt_masktools.dll, v.1.5.x # - mvTools.dll, v.1.11.x # - WarpSharp.dll, v 2003 Code:
# ================================ Code:
# ================================ - gradation curves 1.45 (gradation.vdf) - ColorMill 2.1 (ColorMill.vdf) Filter settings are attached as a VirtualDub .vcf file. |
2 Attachment(s)
Quote:
The only ghosting that isn't due to chroma is due to compression artifacts from being encoded in the wrong mode. The color, and the blocks that you see here are the area where his hand will be in the next field: http://www.digitalfaq.com/forum/atta...1&d=1417366172 http://www.digitalfaq.com/forum/atta...1&d=1417366172 The luma runs at 60Hz while the chroma runs at 30Hz if you ignore the compression changes; in Avisynth you can see this with Bob.VtoY for example. This is one of the few scenarios where ghosting can be visible from future frames rather than echoes of previous ones. Kinda neat... I wonder if motion interpolation to double the frame rate of the chroma could help? Or is that what you're doing? The post-processing fixes you propose are nice, but if merchantlord just wants a quick fix for the combing, I want to again make it clear that he can get results as good as his VLC playback without complex filtering. The quickest would still be seeing what happens with the Oppo set to output 480i. Though now I'm wondering which field order it will output when the source is progressive. If it put outs Bottom Field First, the video in this sample will stutter. |
Thanks for the input, certainly worth considering. If any of the filters are doing motion interpolation, they're doing it without my knowledge. I've seen doom9'ers play with chroma-only in ht emanner you suggest, but they only show cryptic scripts with no explanation -- just copying scripts without knowing something about what they do isn't the best idea IMO.
I guess one could re-set their player for every contingency each time a different problem disc in inserted. Most users today don't even know how to turn their players on and off, and paid somebody $$$ to set it up for them and connect in the first place. A neighbor lady has had her DVD player connected for 5 years but has never used it because she couldn't figure out what to do with the disc. She said the original tech installer showed her how to insert a DVD disc, but she could never remember how to get the disc drawer to open again, so she just left the disc inside 5 years ago and hasn't turned it on since. That story is not only true but is more typical than you'd think. |
Great discussion guys and I'm learning a lot. Sanlyn that tale of yours is a scary one and probably happens a lot more than you think just as you say.
Quote:
|
Does this mean that no one in your congregation, group, or whatever, sees a problem? (I can believe it). Does the ISO maker expect everyone to go into their player menus and alter the settings?
How does this chroma theory explain the motion stutter or low-bitrate artifacts in the speaker's clothes (anyone notice how the lapels and clothing details start to disappear now and then, or see the mosquito noise in that godawful silk fabric's edges?). I've encoded interlaced video myself as progressive without those problems. So something other than re-encoding is afoot. On the other hand, visual illiteracy is an important staple nowadays for the successful use of mass communication and bigger profits. If the audience were indeed discriminating and actually paying attention, most of the garbage that passes for "video" today would be ignored for the junk it is. But why rock the boat? I just don't watch the stuff and don't pay for it. It's too much like 60's science-fiction come true. |
Quote:
As for the popular tendency to settle for second best and to willingly do so (so long as no one tells them that is what they are doing or that is what they must do), that's been going on for far longer than the digital age has been around. It's just that since the digital age has arrived you can really SEE how bad it looks! :) I'll post back when I get to fool around with those settings on my Oppo. |
Quote:
|
Quote:
http://www.digitalfaq.com/forum/imag...nr0p76sp-1.jpg |
Quote:
|
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.