VirtualDub, Histogram and encoding levels

WarbirdVideos · #1 12-09-2019, 03:46 PM

I'm capturing Hi8 using a Pinnacle 710 USB into VirtualDub and encoding with Lagarith. I have a few questions.

1) Consumer grade camcorders expose at 0-235. Video codecs preference to the brightest area of a video signal when recording, and the least amount of preference to darker detail. Do lossless codecs work the same way? Or do they encode video at full range (0-255)?

2) It seems to make sense to boost the gama of dark sections of a video before it's captured to enhance the shadow areas. Technically, this would reduce video noise in dark areas when graded in post. Is this correct?

3) Virtual dub's histogram only shows 16-235. The rest is cut off at either end. The level control (Brightness, Contrast, Hue, Saturation) doesn't allow gama adjustment. "Brightness" moves the entire video signal left and right. "Contrast" expands and contracts the signal equally at both ends. Should I be adjusting the video signal to fit into the 16-235 range equally left to right?

Thanks,
Steve

sanlyn · #2 12-09-2019, 04:52 PM

Video is captured into YUV at 16-235.
Brightness controls black levels (darks).
Contrast controls bright levels (highlights).
The two controls do interact somewhat. You have to jockey back and forth between brightness and contrast to balance the two within the 16-235 range. Darks held within YUV=16 or brighter will protect shadow areas. If you want to tweak various scenes for darks or highlights or midtones, do so after capture. Analog levels change too often during capture for precise adjustment. Adjust brightness/contrast for a worst-case scenario and tweak later.

Your input signal might vary between 0 and 255 depending on your source. Some input sources will exceed that range and clip badly -- which is why you have to adjust to 16-235, whatever your input is sending. You can capture to y=0-255 if you want, but you'll get dark and bright clipping when YUV is expanded into RGB for processing or display.

Lossless codecs accommodate whatever range you've captured.

WarbirdVideos · #3 12-09-2019, 05:48 PM

I'm also boosting the chroma (saturation) before capture. Hi8 is very anemic, and boosting chroma before it's encoded seems to work better. Boosting the chroma after capture looks a lot worse. Seems better to adjust it down if needed than to boost it up after the capture.

One thing I wasn't sure about is if the codecs give more preference to the highlights (50% IRE and above). It seems to matter especially with "noise". I usually tune my cameras to capture more shadow detail these days (flatter image).

I did a test with an unadjusted clip (blacks crushed somewhat in the camera due to lack of in-camera gamma adjustment back in '97), and the same clip adjusted to 16-235 at capture. I could get the unadjusted clip close to the adjusted image, but it took gamma correction, chroma, color balance and level control and just didn't look the same. The properly adjust clip before capture looked better and had less noise in the 15-20% IRE area.

Steve

sanlyn · #4 12-09-2019, 10:03 PM

Making that many corrections during capture seems suspicious to me. There's no way anyone can take your word for it that something looks "better", "good", 'bad", "nice", or whatever. Without a sample of what you're talking about, those descriptions are meaningless. We can't see your videos from here and we can't tell if your monitor is calibrated. Pumped-up chroma that blows away the image in RGB is a common occurrence with those who don't use histograms and vectorscopes to check levels in both YUV and RGB. Your eyes aren't that accurate. I seriously doubt that an entire analog movie would benefit from the same correction from beginning to end. I've never seen analog work that way. It would be news to all us.

WarbirdVideos · #5 12-10-2019, 11:12 AM

I should have mentioned that the videos that I'm working on are "single shot" presentations of veterans telling their stories in front of an audience. The lighting was a combination of fluorescent and 3200k diffused spots. I shot about 40 of these on Hi8 before transitioning into better cameras and tape formats around 2000.

I use a color critical monitor and a Datacolor "spyder" to keep the monitor in tune (although the monitor is extremely close). I'm very familiar with scopes and grading.

I don't mind capturing in the 0-255 range, especially if it will allow for better dynamic range and possibly minimize noise in the dark areas, because all of my videos are graded in post. I just wasn't clear if lossless codecs give preference to brighter areas of the image as digital cameras do.

If a video signal was recorded at 0-235 or 16-235, they will clip at 235 in the camera and won't go below "0" because of the sync pulse area - right? So if the color space is adjusted for 16-235, or even 5-245 before capture, you probably won't exceed specs during capture. The reason for ballparking the levels before capture is to compensate for changes induced by the playback deck and/or the capture card - right?

Thanks for the input!
Steve
PS - I do have a few videos at youtube.com/WarbirdVideos

keaton · #6 12-10-2019, 04:02 PM

The YUV "safe" range of 16-235 and the RGB range of 0-255 has been described in many posts over the years on this forum. Here's a link to a couple of sanlyn posts describing with maybe some different wording that might communicate it slightly different and get the idea across. If you search the forum, you'll probably find more.

Read the last paragraph of this post http://www.digitalfaq.com/forum/vide...html#post47840

Here's another example http://www.digitalfaq.com/forum/vide...html#post52169

Basically, what is trying to be communicated is the difference in what are considered legal video levels (i.e. in gamut) in two different video "domains". There is the YUV domain of the TV world, and there is the RGB domain of the PC world. When we capture video to PC, i.e. RGB domain, the video levels that come from the YUV analog world are converted to a digital RGB world. The mapping of anything at or below 16 YUV goes to 0 in RGB, and anything at or above 235 in YUV world goes to 255 in RGB world. Therefore, the preview histogram in a program like Virtualdub shows if your YUV is in the safe 16-235 range by showing anything below 16 or above 235 as red in the Virtualdub capture histogram. See http://www.digitalfaq.com/forum/vide...html#post45238 If you have any red in the histogram, it is YUV that will be clipped/crushed to the same levels of 0 or 255 when working in the RGB domain.

A program such as Virtualdub works entirely in RGB domain. However, you can capture to and also save files to lossless codecs from that Raw RGB to YUV based codecs such as HuffYuv and Lagarith. When viewing a YUV histogram via an avisynth script command on captured video, it shows the legal range of 16 to 235 to equate things to the YUV domain. So any 16 to 235 based histogram (likely an avisynth based one) is not saying 16 to 235 RGB, it is doing the translation for you since it displays Y, U, and V channel information. If a histogram shows things in terms of R, G, and B channels (likely a Virtualdub based one), then it would show a 0 to 255 legal range. I am referring to Virtualdub histograms used in post work, not the Preview Histogram mentioned above. That one is 0 to 255, but is really showing you things from a Y channel perspective. Digital video is not often stored in Raw RGB file format, because it is too large, and can be losslessly compressed to the aforementioned codecs. But RGB is the intermediary though which Virtualdub works when reading from a codec or converting back to a codec that has a YUV based representation.

I hope this helps get the idea across. I know it took me some time to truly understand the concept of TV vs PC or YUV vs RGB. You can either use an external hardware proc amp in your video chain to get levels "legal" or use the internal proc amp on your capture device (if it has one) to get these levels legal. If you want to give yourself some "headroom" by capturing with YUV levels even further in than 16 and 235, then you have some freedom with adjustments in post to make shadow or highlight adjustments to utilize the full gamut. Although, I wouldn't suggest too much headroom. Just enough to be able to not have your tools crush your video when you try manipulating it. Avisynth has a great command named levels that can help with remapping to the full gamut. Many forum posts from sanlyn and probably others on the Levels command and how it works, and using other histograms or readouts available in avisynth to determine what the min and max levels of your captured video is. Those values are useful inputs to the Levels command.

Regarding the anemic color values of camcorder video. I've had a bit of experience there as well. I was quite pleased with what I could do in post to correct and boost the color without going overboard using tools such as Virtualdub and Avisynth. There's a wealth of forum posts, once again many from sanlyn, that show what can be done with these tools to identify and correct color/contrast problems using RGB and/or Hue/Saturation based tools, as well as amplify the corrected colors to bring the video back to life. I've also learned quite a lot from Alexis Van Hurkman's Color Correction Handbook. It is well revered for both beginners and advanced colorists.

Best of luck to you.

WarbirdVideos · #7 12-10-2019, 04:03 PM

I ran another test and it is mandatory that you capture at 16-235 in Virtualdub because Lagarith stores the signal as 0-255.

WarbirdVideos · #8 12-10-2019, 07:38 PM

Thanks Sanlyn & Keaton. It's making sense.

Is there a way to put a link to detailed information such as Keaton's above. I know that Sanlyn has a ton of info as do others, but in this case, a link under "Guides" entitled "Yuv 16-235 & RGB 0-255 explained" would be helpful. Newbies come here and ask the same questions over and over again and one gets lost in the threads. Or have I missed where this info is stored?