Huffyuv capture questions
Hi Everyone
Please forgive the newbe questions. I've never really digitised any VHS footage before with any seriousness until recently. I have a Hauppauge USB-Live 2. I captured my VHS tapes using a Panasonic 930 SVHS machine via a AVT TBC using VirtualDUB and HuffYUV (single thread). I had to use the Windows WDM interfaces as for some reason, DirectShow wouldn't work for my device. I have some questions that I'd appreciate some help with: 1) Is using the WDM device any different from DirectShow from a quality/capture standpoint? 2) I use 25fps to capture my PAL VHS tapes. Is this correct? I'm aware that VHS-PAL is 50i 3) Are my recordings still interlaced? I see some "combing" artifact during playback, and I'm not sure where this de-interlacing happens 4) I used AVStoDVD to make a DVD from these captures. Is my DVD interlaced or progressive? I appreciate your help Thanks |
Quote:
Quote:
Quote:
Quote:
Quote:
Mount your AVI capture or a VOB in VirtualDub. Add the deinterlace filter (use the yadif version available in the deinterlace filter's dialog window, using both fields for double frame-rate playback). After enabling the deinterlacer, play the video one frame at a time using the VirtualDub play controls, and observe motion differences between the input window and the output window. If there is motion in an interlaced frame, it will take two clicks of the frame play button to show the complete motion in the output window. If you don't know how to use VirtualDub, you should learn to use it as a quick and handy analysis tool. If you like, you can post about 8 seconds of your original, unaltered huffyuv capture directly into the forum. Choose a short sequence that has motion of some kind (someone moving, camera pan, etc.). Cut the edit in VirtualDub and save it using "Video" -> "direct stream copy" mode to avoid altering the colorspace or compression. A few seconds of huffyuv YUY2 would be within the 99MB posting limit. Please do not post YouTube samples -- they are never interlaced and are always badly re-processed. |
Thanks, I'll upload a small clip to show you.
In the meantime, I'd be grateful if you could tell me if I'm correct/wrong with these: 1) Interlaced digital video is never designed to be distributed to average users for viewing on a computer. It is only really designed to be displayed on a TV where interlaced display is native (CRT) or has proper de-interlacing hardware (LCD TVs). 2) From a digital perspective, interlaced and progressive video is actually the same thing as far as the computer is concerned. All "deinterlacing" does is applies a filter to blend the combing effects (created by half of a frame being from a different moment in time) by predicting/analysing movement, for better viewing on a progressive display. 3) When viewing TV broadcasts (DVB in Europe, ATSC in the US), the "i" or "p" flag set in the stream just tells the TV to do it's blending or not as above. Is that correct, or am I totally off? |
Quote:
By and large, movie makers don't create films specifically for computer display except in some cases where a program has to be modified for the display characteristics of computer monitors, which differ from the display characteristics of TVs and projection devices. With most ordinary sources, interlaced or telecined vbideo is deintrlaced by removcing alternate frames for the chaopo online media players that can't deinterlace. This throws away half the spatial and temporal resolution of the original. The alternative is double frame rate deinterlace retaining all frames, but streamiung at twice the frame rate at normal frame size is a different proposotion. You would have to ask movie makers why they don't produce films specifically for TV screens that display at 24, 25, 48, 50, or 59.94 Hz. Quote:
Quote:
|
Hi Sanlyn
Thank for your extensive reply. I've still got some questions but I really appreciate you taking the time to explain this to me. Regarding me saying that interlaced vs progressive video is the same in the digital realm, perhaps I wasn't articulating my thoughts correctly. What I mean is, is that 25fps is 25fps period. The only difference between interlaced vs progressive, is that for a given frame, interlaced video has half of the rows from one instant in time and half another instant; whereas progressive has all rows from the same instant. Is this correct? Regarding my de-interlace description, perhaps blending gives a poor result, but it's still one (albeit crappy) way of deinterlacing, right? I now understand that better methods are like the double-frame Yadif which split each field in to it's own frame. Am I correct in saying that the Yadif algorithm fills in the blank rows using some cleaver magicness? I assume this to be the case, as a field in interlaced video is only on alternate lines, so the other lines must be filled in somehow, right? Or does Yadif just simply stretch the field to fill the frame? And now my final question: In computer video, is there a field in the metadata somewhere that tells the player if it's interlaced or progressive? I'm assuming not, as VLC needs to be explicitly told, right? I also can't find a way to tell if a video is interlaced or not without going frame by frame in Virtualdub. Thanks! |
Quote:
You are also aggregating standard definition algorithms with HD algorithms. HD formats have provision for double-frame rate video as well as for non-telecined film speed video, and interlaced/telecined formats as well. https://www.videohelp.com/hd#tech. Meanwhile the internet is strictly progressive. Quote:
There are better deintelacers, such as NNEDI3 and QTGMC, depending on the effects you want, but they aren't real-time deinterlacers. QTGMC probably does a better job than any playback device these days, and includes some very good (as well as occasionally destructive) denoising and repair options that can be enabled or disabled. QTGMC's resizing and motion interpolation methods can be very sophisticated. But interlaced originals are usually archived as interlaced because future developments could eventually produce better deinterlacing with fewer side effects and greater speed. Quote:
If you see interlace combing, you're looking at video that was interlaced, is interlaced, or could be telecined, or was telecined at one time and badly processed. If you find it too difficult to perform simple tasks with something like VirtualDub, I hasten to say that the usual NLE would be too advanced, a "pro" level app such as the over-hyped Premier PRO or Vegas would be a stern challenge (and neither deinterlaces that well to begin with), and Avisynth would be out of the question. Virtualdub offers many easy and quick ways to analyze the frame structure of many video formats and will accept more formats than simple AVI. VirtualDub can also offer some sophisticated color and levels analysis tools. Avisynth can do that as well as break down video by row and/or column, by color channel in YUV or RGB, and pull off a lot of other tricks. Usually you'd need only a simple subset of the available tools from either app. Here's an old guide in .htm format: Neuron2_How To Analyze Video Frame Structure.zip. |
Thanks for your reply!
I'm not having any problems using virtualdub to deinterlace. It's working great - I'm just trying to understand how it works. My blanks frames remark is directly related to your following statement: Quote:
|
1 Attachment(s)
I'm uncertain about which methods various deintelacers use to interpolate frame sixes. Some are simple algorithms, others are complex motion-compensated interpolations. Most Avisynth and VirtualDub third -party filters are accompanied by their source code, which is usually C++ or assembler language or both. You will malso find more details in development threads on doom9 that are detailed discussions on the use and innards of AVisynth filters. I am not a programmer of filters, I just study how experts use them.
Attached is an avi with four side-by-side sections of the results of deinterlacing the same difficult video with four different deinterlacers, from left to right: yadif, MVBobMod, TempGaussMC, and TDeint/NNEDI. The sample shows how different deinterlacers handle different noise and object rendering problems. TempGaussMC (aka "TGMC") is the predecessor of today's QTGMC and is still in use. |
Is there anything specific to look at in that clip? They look largely the same. The outside two appear to bounce up and down a little, and BobMod seems a little blurrier?
Could you throw a field blend in there with the same clip? |
I wouldn't be able to modify that test avi because I don't have the original. And if I did I wouldn't throw a field blend into an interlaced original. It's easy enough to do -- encode an interlaced video as progressive or use Premiere's field blend option. Or resize a video horizontally without deinterlacing first, then restore the original frame dimensions.
One purpose of the test avi looks specifically at line twitter, aliasing, and moire created by various types of objects and interpolation methods. It's a shame you can't see that the TempGaussMC panel is far more stable and cleaner than the other three, and everyone has known for a long time that simple bob() or its derivatives are blurrier than other methods. But in fact all of the methods shown except TGMC are in wide use in one form or another in various desktop apps and in some Avisynth published scripts, and each has their particular uses and purpose. But it's beyond this discussion to get into field research on every aspect of their use, and is beside the point. The post and discussion appeared about 2008 in doom9 and elsewhere and the original has since been lost to the ages. Originally the discvuissions dealt with whether or not faster deintelace speeds were worth the results, with TGMC being the slowest and yadif and simple bob() algorithms the fastest. The title of this thread no longer has anything to do with the discussion. However, I did neglect to mention one issues from an earlier post: Quote:
There are other differences in the digital domain. PAL video is often made from 23.976 or 24fps film originals. Various means are used to bring film speeds up to 25fps, including just speeding things up a bit (I hate that) or using various forms of pulldown (aka telecine). Some frames are configured into fields and interlaced periodically so that certain fields periodically repeat at certain intervals, or in some cases duplicate frames are inserted at intervals and played at an accelerated frame rate. In the restoration of silent films shot at anywhere from 12 to 20 fps, various duplication and telecine techniques are used to create the impression of 25fps playback. The same thing is done for NTSC, and for PAL to NTSC conversions. So, in many cases, "25fps" isn't nearly as straightfoward as it seems. |
Quote:
Quote:
|
I once looked for didee's original for 3 hours. No wonder I didn't find it. Thanks for both notes.
|
Quote:
A 1 second interlaced or progressive clip will both have 25 frames. Surely, the "different image every 1/50 second" is down to how the computer handles the playback/processing of the video, rather than the video itself. Why this mundane issue is important in my head, is that previously, I always though that interlaced video was of a totally different "format" (analog codec, if you will) to a progressive video. As in, fitting a circle in to a square hole. I always thought it was impossible to play back an interlaced video on a progressive screen without any form of deinterlacing. But it turns out to not really be the case. It's just lots of pictures (25 a second) with alternating lines of each picture being from a different moment in time. |
I can't guess what would make anyone think that interlace was a strictly "analog" phenomenon. Today's digital broadcasts are interlaced, so is standard-definition interlaced/telecined digital DVD, so is high-definition BluRay at 25 or 30 fps, and so is PAL or NTSC digital DV tape as well as many of today's digital HD cameras and webcams. I don't know what video forums you've been frequenting, but I'd advise taking some of them with a hefty grain of salt. You might also want to consider various pulldown or telecine schemes, which are a mix of progressive and interlaced frames, along with some field duplication. Some pictures of various forms of pulldown for PAL and NTSC from progressive film sources are shown at https://en.wikipedia.org/wiki/Telecine.
Unfortunately you can't the internet. I saw a website that defined "telecine" as a way of bootlegging retail videos. Now, how do you suppose anyone came up with that line of total nonsense and B.S.? Sadly, many naive readers would believe it. |
I never said it was just analog. Nor do I think it is. All I was saying was that I used to think they were totally different formats and that it was impossible to display interlaced video on a progressive display (without deinterlacing) when in reality it's just what a frame consists of. Very simple.
Although, I will admit, that since CRTs natively scan in such a way that tends well to interlaced sources, this is quite possibly where my confusion stemmed from. Btw, in the UK, we have free OTA digital broadcasts in 1080p now :) |
Quote:
Quote:
The technical structure of compressed digital video can differ depending on whether the data has been flagged as interlaced or progressive. Huffyuv, Ut Video, and others have thresholds or checkboxes to tell the codec you're compressing interlaced. This allows the codec to correctly compare across separated fields (half-height frames, 2x FPS) rather than trying to compare combed images for better compression ratio. Note that this flag is used internal to the codec only. The AVI container has no standard interlacing flag and so you must manually tell a player to deinterlace your captures to view them properly: http://www.digitalfaq.com/forum/news...nterlaced.html In both analog and digital, it's nonetheless possible to send progressive images through an interlaced "container". |
Looks like I'm learning even more new things :)
Quote:
I believe this to be the case, as you are able to play back an interlaced HuffYUV video without deinterlacing enabled. (It just looks combed) Quote:
Thanks! |
Quote:
Quote:
Quote:
Quote:
480i CRTs can display 240p60 but not 480p60. (If they could do 480p60, interlacing never would have been invented.) 1080i CRTs can display 540p60 but not 720p60. They have to internally downscale 720p using digital processing before presenting it. |
Quote:
Quote:
Thanks for taking the time to answer my questions. I know I'm asking really weird questions, but I'm so interested in how all of this works. The fact that I was able to view non-deinterlaced interlaced video on my PC blew my mind. |
Quote:
Quote:
Probably better off posting this in my own thread but the main reason for my interest, beyond my general unhealthy desire to know absolutely everything about absolutely everything, is I'm looking at encoding my restorations as progressive. At the same time, no point trying to copy across half this thread and jt_retro might get something out of it. |
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.