Hi all,

I have a video from a friend of mine that has a significant red overcast for most of the video that I captured. It is so prevalent that I double checked that I had not accidentally mis-set the hue or contrast ranges.

In looking for approaches others have used to correct this problem. Searches I have done are for color correction and chroma bleed. The reason I searched for chroma bleed is when I look at the right edge of the video, there seemed to be a 10px or so wide band that looks like it might be correct so I thought I might have a shift.

I have made a couple of attempts, but the results seem very flat. The video IS in a church and many churches have really poor lighting so flat may be the best possible outcome.

First, attempt, I tried just using ColorYUV and bringing the red levels down, managing the yellow level some and adjusting the overall luma to lighten up some of the darker areas. I had read other examples where folks had looked at the overall RGB luminance values and could tell that one of the color channels ( the one I saw was the blue channel was wiped out) was very different from the other two.

Second attempt is where I saw there a mask was used to preserve the luminance values and then adjust the bright and dark sections of the image independently. My understanding here may be off though. I would run the script and look at each of the layer prior to the Overlay() and to my eyes it did not seem like the dark layer was doing much anything.

I'll just include the inner section where I swapped out the vanilla ColorYUV() with the masking operation:

First, am I thinking of the problem here in the right way? I'm very new to processing video so I feel that I'm near the bottom of a moderately steep hill.

Second, I feel like my first approach on using ColorYUV() was very imprecise. I was looking at the YUV histograms and adjusting values to keep the majority of the signal it between the two colors ( ala between the red and green or between yellow and blue ). If the best tool the vectorscope in these situations?

Last, and not directly related to the color issues is that I included a Trim() early in the script, but when I processed it through VirtualDub with the AVISynth script as the source, it didn't truncate the video file like I expected it to. Do I need to approach that differently?

Note: I do have video examples, but it seemed to be having challenges letting me attach files...all three clips are < 99MB so I'm not sure what's going on there. You can see where I'm starting from with the screen shot. Any suggestions on how I can get the clips attached would also be appreciated!

-- merged --

Here's the first script output

Kind of makes me wonder if that is a multigeneration tape and not the original recording where errors may have stacked up between generations? A hardware proc amp could be used during the capture, but I'm not totally sure that it would be any better than correcting color after the capture or not unless the red is so intense that it is clipping. I assume it plays that same way on multiple VCRs?

Original Capture and the result of the Mask script.

Is it the original tape or a copy?

What VCR, TBC, and capture card are you using?

It's an original tape as far as I know. I don't have the background on who did the camera work.

JVC SR-MV45 -> TBC-1000 -> ATI AIW.

Vast majority of my captures look great.

Yeah, that is one thing I checked is that it was playing the same way. Tried a Samsung and Toshiba VCR and both actually had the significant red overcast.

Like I said in the other reply, I don't know if this is original tape or not. I'll certainly have to ask

Here's a thread from a few years back where sanlyn and I each took a shot at a video with some very strong color problems. https://www.digitalfaq.com/forum/vid...lor-spots.html It assumes Virtualdub and Avisynth. I prefer sanlyn's method.

In that thread, I used a Virtualdub filter for hue/saturation adjustments. Nowadays, I use Tweak, as you'll see in sanlyn's post.

From what you describe, I'd guess that on a vectorscope you would see some very strong readings on the vectorscope, and some possibly clipped outside of legal gamut. Tweak's selective hue and satuation adjustments can really help reduce extreme problems. I've also used ColorYUV to try fixing gross color channel problems. RGB adjustments (in Virtualdub filters) may be more helpful for smaller adjustments once the major color problems are addressed.

Also, there's a reference to the Color Correction Handbook by Alexis Van Hurkman, if you really want to dive into the topic of color correction. Although, as I've learned with VHS (especially camcorders), there's only so much you can do to correct color because it's not so color accurate or consistent to begin with. You may have to break it down into a lot of smaller clips if you really want to make each shot color accurate. Which I have done before for video I thought was worth the effort. Also, have a fairly color accurate monitor if you get into this kind of stuff.

If you search this forum for ChromaShift (http://avisynth.nl/index.php/ChromaShift) you should see some examples of how it can address color bleed. Although, I've heard it's not in 64-bit versions of Avisynth, and it's been replaced by ChromaShiftSP

Thank you for the pointers to those links! I had found a few others that Sanlyn had contibuted to and benefited from his explanations. Somehow I had missed that particular thread. I have homework to do!

Sorry I don't have the time to dig into this specific clip. I did preview it, and it does look like you've got your work cut out for you. With so much oversaturation to start with at so many different hues, this could be quite a challenge. The techniques of targeted saturation reduction at certain hues, or possibly hue adjustments by using Tweak's hue parameter to rotate some or all colors on the color wheel are what can typically be done to address large color problems. However, you may find it difficult to re-amplify the color after corrections because you cannot totally remove the color errors. I'd suppose it's better to live with more muted colors than amplified colors that are obviously not right.

Best of luck to you.

Ha! I appreciate the candor and the pointers. That's exactly what I was looking for is direction/correction in how I was thinking.

Maybe better throw away the coloring and use something like deoldify/ddcolor/DeepEx to invent the colors
(results could be improved if there are reference images; also applied ReduceFlicker)

That looks significantly better than what I was reaching. I will certainly take a look at those.

Is that in Hybrid yet? :D

I need to try this, new to me.
In fact, is that a new project entirely? I didn't see it 4 months ago, when I was looking into non-manual "AI" video colorization.

Yes, has been in Hybrids torch-addon since March. (did improve a bit since then, but has been there for a while)
For general info what happened, read: https://forum.selur.net/thread-3595.html

Cu Selur

The other thing is the (what I can only describe as) extreme motion blur or "motion trails". I don't think I've seen any other tape having it to that to that degree. The only time I've seen even mild motion blur get introduced was when using a Blackmagic Teranex 2D to do deinterlacing/upscaling via hardware, but it was to a much lesser degree. Almost makes me wonder if that was some sort of a "visual effect" enabled on the camera that was recording it at the time.

In some instances in the whole video it's very noticeable and in others it's not. There is also a few spots where it looks like they swapped a filter on or off.