DV --> PC color correct --> DV

[joshbm]

Hey all,

I recently put on a small clip onto my hard-drive of DV (Digital Hi-8) footage. I then dragged it into Media Studio Pro 6.5 and added a color correction to test the results. I results look great! I get way better looking video. No bleeched out white video. I then proceeded to put it on DV again (I do this to all my final projects) and it looks horrible!! Why is this and how can I fix it?

Also if it looks good on the PC, shouldn't it look good on DVD?

Thanks,
Josh Bm

[bb]

Welcome to the forum!

Which settings did you use for the color correction?

bb

[joshbm]

Thanks for the welcome! I thought no one would notice lol. Well I used this filter:

"Highlight"

Red
Highlight 3
Midtone 42
Shadow -29

Green
Highlight 8
Midtone -13
Shadow -6

Blue
Highlight 7
Midtone -22
Shadow 12

I could give you a before and after pic if you like so you could see it visually.

[joshbm]

Oh and also I haven't yet tried but I also add a Saturation to get more crisp colors. I don't know how it would look putting it back on DV though.

[Vulcan]

When you say it looks awful when put on DV do you mean when the DV is viewed back on PC or TV? If it is the latter it is most likely explained by the fact that "the colors you see on the PC are not the colors you will see on TV."

I haven't tried color correction my self since I'm yet to buy a small TV that I can hook up to my NLE/computer so that I can view the result as I do the process. There is a tutorial/explanation made by BillyBoy that refers to Vegas (which I use). The tutorial can be found here: http://www.wideopenwest.com/%7Ewvg/tutorial-menu.htm

[joshbm]

It looks not the greatest on DV as well as the TV monitor. The colors are not as rich as they were on the monitor.

[Mug Funky]

DV is a bit of a naff format for colour correction, due to it's being designed principally for TV viewing.

TVs are known to have problems displaying colours that are too saturated, and luminances above 235 and below 16. so when DV is recorded it tends to be in the TV viewing range (unless your camera can handle super-white, where it uses the full available brightness).

the problem occurs when the DV is processed on the computer. some decoders brighten the input DV (so it looks much better than it actually is), and some don't. when footage is recompressed after being "corrected" by one of these decoders, you will have luminance and chrominance values far in excess of what looks good on a TV.

on the other hand you could have a decoder that leaves the colours in their native 16-235 range (called CCIR, generally). they will look duller and greyer, but actually preserve more information. if you correct this dullness manually with a levels filter or what have you, the output will again be oversaturated.

without knowing what you are using to decode your DV (mainconcept DV codec, and adaptec corrects the range, where canopus and pinnacle do not), or how you are displaying, etc, there's not much advice i can give beyond the above. (that should probably help you though)

[FredThompson]

If you're using NTSC DV, be aware that it uses 4:1:1 color so screws things up. Samples on my site: http://www.geocities.com/fredthompson6

Xesdeeni has a saturation filter to help correct this. I'm just now testing it.

[joshbm]

Thanks for the replies guys!

@ FredThompson:

Yeah I'm using NTSC DV 720x480...
So where's Xesdeeni's plugin?

@ Mug Funky:

Thanks for the info. Maybe if I told you what I use you would know of a fix or a decoder that simply leaves colors in the native 16-235 CCIR range.

I am using Ulead's Video Studio 6 to take it off and I have been lately going straight to MPEG-2 to save on much room and still have good picture.

[FredThompson]

range limit is easy with AviSynth.

Hang on. I've been doing some work making screengrabs.

Check here in an hour: http://www.geocities.com/fredthompson6

xesdeeni's filter is in this thread: http://forum.doom9.org/showthread.php?s=&threadid=57013

I'm almost done working it into my cleanup script. Got a little sidetracked with the wavelet denoiser. Man, that's nice. Very nice. Nicer than anything else I've ever used. Unfortunately, it's dog slow, around 2-3 fps on my Athlon 2100. By way of comparison, my script runs around 9-10 without Xesdeeni's plugin, 7-8 with it.

The result is darn nice. Be aware, however, that you can't save in the DV format without a lot of loss.

MainConcept has range limiting as a setting if you wish.

[yg1968]

Mug Funky,

I gather form your response that what you are saying is that there is no point in correcting the colours of DV if your goal is to watch the edited footage on a TV. The only reason to correct the colours would be if your goal is to watch the footage on a computer (for example in dvix). On another point, I am assuming that most editing software will only render footage that is modified (for example transitions). In other words, the codec should only affect the edited parts of the movie (i.e. the footage with the transitions, titles, filters, etc). I mostly use the Microsoft Directshow built-in (type 1) DV codec and haven't noticed this colour washout problem.

[FredThompson]

No, I meant you should be aware that if you have NTSC DV and you process then save back to NTSC DV you have the very real potential for image degradation because NTSC DV is 4:1:1

[yg1968]

From what I have read 4:1:1 is not that bad. 4:2:0 is considered better for PAL colours but not for NTSC.

See the folowing article which states that the difference with RGB cannot be seen by the human eye.

See points 2.2 and 2.4 of the first article which states:

2.3 Color-Space Conversion

The first transition made from the raw data collected by the three CCDs is from the "RGB Color-Space" to the "YIQ Color-Space". This is primarily done for compatibility issues with NTSC broadcasting standards as well as achieving what is considered "visually-lossless" lossy compression. This seemingly inconsistent phrase refers to the fact that the conversion between these two color spaces results in an unrecoverable loss of color detail, but because of the nature of human visual perception, this reduction in color resolution is unnoticeable. [...]

2.4 Down-Sampling YIQ
[...]

The end result is that 4:1:1 systems have 25% less color information as 4:2:2 systems. To add some perspective, the resulting quality loss of either system of data reduction is still very negligible to the perceived image quality. The two formats at this stage would be nearly impossible to discern with the human eye.
See these articles:
http://www.rtproductions.net/home/documents/DVCAM.htm
http://www.rtproductions.net/home/documents/Executive%20Summary.htm

See also these related articles
http://www.adamwilt.com/DV-FAQ-tech.html#colorSampling
http://www.lafcpug.org/feature_capture_card.html
http://www.adamwilt.com/pix-sampling.html
http://www.m2w.net/reviews/dvcompression9808.html

[FredThompson]

Those look like good references. Thanks.

Having played around some more I think there are some codec issues. That gets tossed around here frequently with opinions stated and no real comparison data.

The sticky about codecs claims the Panasonic is more accurate yet does not offer a true metric, just an opinion. The statement is made about the result of the Panasonic codec when looking at a home-brew sample. There are just too many variables in that process to know for sure what is going on.

I've run into some problems with converting from DV to another colorspace resulting in noticeable loss of detail.

Because of this, I think a true comparison is needed. So...I'm going to halt all the other things I was doing and work on this. It needs to be done anyway because the first thing that would be done with a DV capture is some form of colorpsace conversion or modification. I've got a Snell & Wilcox test sample which can be used but it's not originally DV. Thus, we have the potential for corruption during an encode process. IOW, there needs to be a DV sample with known good status.

Even if we only test 4 codecs, we'd have to encode with each and decode with each to yield 16 samples. If we have a good reference capture and a torture frame from camcorder (I have a good one), we're up to 32 samples. At 6 codecs we're up to 64 samples (6x6x2.) It gets even more complex as the MainConcept codec has 3 encoder options and 5 decoder options that will affect results. I'll attempt to get a response from them but it's already the afternoon in the U.S. so they're probably closed in Germany.

The point being, there needs to be some sanity in the comparison or we'll drown in data.

The only other comparison I've seen looked at multi-generational use of codecs to reveal color problems. The M$ codec had a notorious preference to make everything green. I have that complete site mirrored. The codecs are outdated but it is interesting.

If, after thinking about what I just described, you have suggestions about how to properly make a comparison, please let me know. The last thing I want to do is end up with 1000+ samples from which to try to find the most accurate.