Thanks for the additional capture info.
High levels of interlace noise are not unusual. In this particular case the excessive combing is mostly a result of the properties of the original camera and recording. It's made worse by equally excessive camera motion, which ultimately eats up bitrate and accentuates residual noise. It's one of many problems with with amateur video. You can't blame the player or video editors for consumer error.
Another common problem with beginning captures is invalid luminance levels. Sometimes this can be repaired, sometimes it can't. In this case the camera helped somewhat because the light levels being photographed exceeded the camera's capacity to record some of the overly bright data, resulting in some clipping of brights at the outset. During capture there was overly bright material that managed to survive the camera's exposure system, but the brights still exceed valid YUV digital video levels.
The image below is a YUV levels histogram of a typical frame from your second sample:
The
#1 arrow in the histogram's upper-right corner points to out-of-spec bright values that exceed the safe video range of 16-235. The portion that exceeds the acceptable range is shown in the right-hand shaded border of the top (white) band. The histogram measures a brightness range exceeding 255. YUV values of 16-235 are expanded in RGB to 0-255. If brights in YUV already exceed 255, an RGB 0-255 display (such as a monitor or TV) cannot accept those values. So excess data values are clipped (destroyed). In a broadcast system, out-of-spec YUV values are rejected. Those values will appear distorted and can damage pro systems.
The
#2 arrow in the above histogram shows that your capture device will not accept black values lower than y=16. Black clipping occurs before the signal gets into
VirtualDub. In this case the clipping doesn't create much harm, but in an underexposed, night-time or interior shot the clipping would affect the dynamic contrast range of the image.
When I examined the lines-on-pan.avi sample I used Avisynth functions to correct the luminance range and contract YUV values into y=16-235, thus retrieving some of the bruight data and avoiding loss when the video is converted to RGB for display. The Avisymnth coded I used was:
Code:
ColorYUV(cont_Y=-15,off_y=-8)
Levels(16,1.0,255,16,250,dither=true,coring=false)
The resulting histogram is shown below, with all values within the y=16-235 safe zone:
In the same sample there appear to be one or more missing frames between frames 61 and 62. I was unable to locate the same problem in the larger off-site .avi, but the larger sample was unworkable anyway because of strong luminance clipping in RGB.
Quote:
Originally Posted by puleddu
I'm using the following setup:
Panasonic NV-FS200 -> [ Panasonic DMR-ES10/15 ] -> ATI HD 600 USB -> VirtualDub 1.9.1 on Window XP Pro SP3
The DMR-ES10 and -15 (I tried both) I put between brackets because I also tried not to have them in the chain. I didn't do to troubleshoot this issue, as I didn't think it could be the case, but to check if I'd spot any differences in general; I didn't notice any.
|
Those elements aren't related to the combing effects. However, those effects and other glitches would look much worse with lesser gear. Your setup should produce very good results, far beyond average. The NV-FS200 does have a line tbc, but you might still need the pass-thru to help with outlaw tapes that refuse to track cleanly. The ES10 and Es15 aren't completely transparent but they do help to make outlaw tapes look smoother. I have an ES10 and ES15 myself and have posted samples of their use.
Quote:
Originally Posted by puleddu
the property Video Format › Video Standard under the Capture pin... menu, reads NTSC_M. It sounds off, considering I'm in a PAL Country and I'm using PAL equipment.
|
Match the video standard setting to the tape's format. Your VCR does play NTSC properly.
My basic approach for filtering the sample was to deinterlace with QTGMC to producve a 59.94fps progressive video that I saved as a YV12 lossless Lagarith file for further processing into various output formats. Note that deinterlacing, even with QTGMC, is a destructive process and isn't always required for cleanup or color work. In this case I first used SeparateFields() to clean up some edge halos and slightly fuzzy edge ringing, something you'd see with just about any tape on just about any VCR but which are very mild with your Panasonic. I followed that with QTGMC and a couple of filters to smooth combing and buzzy edge effects (vInverse2 and santiag). One side effect of QTGMC's motion compensated algorithms is that it does tend to ameliorate some source interlace shimmer and other problems. It isn't perfect, but it's better than any other deinterlacer.
From this 59.94fps progressive file (I refer to it as the "60p") I was able to produce the following:
a) I opened the 60p video in Avisynth and reinterelaced the progressive frames back into the original 29.97fps interlaced format. This was done for producing a standard DVD final delivery version. The result is attached as
Lines_On_Pan_29.72i_standard_interlace_DVD.mpg. You'll see that playback is cleaner than the original, but excessive camera motion really works against us. Combing on erratic motion isn't altogether subdued, although it looks smoother on TV than on a PC. How it fits in with the rest of your movie depends on other segments that don't have so much frantic motion.
b) One method for smoothing the excess combing effect is to avoid interlace altogether -- although there are negatives attached to that. First, if you want DVD or standard def BluRay for output, you can't use progressive video or double-rtate 59.94fps. You don't want to upscale to double-rate 720p because that never looks good with low-resolution VHS and is a dead giveaway that a newbie did it -- the average HDTV can upscale much better than you can.
But a common trick for getting this 59.97fps video into DVD or SD-BluRay format is to discard alternate frames. That restores the 29.97 frame rate. The next trick is to tell your MPG or Bluray encoder to add interlace flags to the final encode, thus tricking your players into thinking your video is interlaced (many of them would play it as interlaced anyway). Besides, if this trick is good enough for the pro mastering labs it's good enough for us.
I opened the "60p" video in Avisynth and discarded all odd-numbered frames to get 29.97fps playback. Then I told my encoder to encode it as 29.97fps with embedded interlace flags and a DAR (Display Aspect Ratio) of 4:3. The big negative here is that the method discards 50% of the movie's temporal resolution. Some of the motion won't look as smooth as the interlaced version. On some playback systems this might not be seriously noticeable, on systems that aren't so smooth to begin with you might see some motion judder with all those kamikaze camera movements. But there are no combing remnants. The result is attached as
Lines_On_Pan_29.97p_interlace_flags_DVD.mpg.
c) For PC or network display or for internet mounting, I opened the 60p file and resized the frames to square-pixel 640x480 (which is a 4:3 frame size). The actual playback speed is 59.94fps progressive. Audio was reduced to the typical internet rate of 44.1k. The result is attached as
Lines_On_Pan_60p_4x3.mp4.
d) Alas, some setups don't make so nice with 59.94fps video even at standard definition and will balk at playing it. And some websites are still slowpokes about that frame rate. So I opened the 60p file, discarded odd-numbered frames, resized to 640x480 using Spline36resize, and encoded a 29.97fps square-pixel video. The result is attached as
Lines_On_Pan_29.97p_4x3.mp4.
All of this work was done in Avisynth. It might be possible to do it in VirtualDub, but the results won't look the same. If you need some Avisynth help or details, ask here. Meanwhile the same filters might not work for other segments, but we only have only one short one to work with, so if you feel that posting more samples would be helpful, go right ahead.