Help Matching Levels Between 2 Videos

[EpheMeroN]

Video1 (images on left) and Video2 (images on right) are both from the same televised event. However they were recorded from different sources and have different levels, and color, etc. Video1 was a direct-stream-rip saved to .ts format, and Video2 was a straight-to-xvid capture. I need help matching Video2 with how Video1 looks (which is the video that I believe needs no correction but correct me if I'm wrong).


http://img225.imageshack.us/img225/8673/video1ei5.png http://img65.imageshack.us/img65/8506/video2pw0.png


http://img68.imageshack.us/img68/791/video10nv4.png http://img68.imageshack.us/img68/6090/video20mi4.png

[Terranigma]

I know this may sound weird, but have you considered using tdeint's clip2 feature? It allows you to swap frames from an external clip. Don't know if that's what you want to do, and judging from the pics, you might have to resize the .ts source to match the xvid source.

[EpheMeroN]

I know this may sound weird, but have you considered using tdeint's clip2 feature? It allows you to swap frames from an external clip. Don't know if that's what you want to do, and judging from the pics, you might have to resize the .ts source to match the xvid source.
The only reason why I am using these 2 videos of the same source is because the main capture (Video1, the .ts capture) ran out of hard drive space 30 minutes before the event was over. So what I will be doing is making a DVD out of this by merging what was missing with the video from Video2 (the XviD capture). Since it'll be a DVD I will be resizing to a dvd-compliant res when I encode it. I just want to ensure both videos have the same / proper levels and colors.

[Alain2]

Levels don't seem different very different to me from these pictures ; here (http://img302.imageshack.us/img302/1472/tempcs1.jpg) is what histogram says from your pictures. I think it's more that xvid was too low on the bitrate and blurred everything.. (like all the red dress details lost)

[Richard Berg]

Agreed. Add some spatial smoothing to #1, boost the contrast on #2, and they'll look very similar.

[Revgen]

I think mg262 came out with a filter called ColorLike. It was supposed to do what you suggested by comparing clips and automatically making the color of one resemble the other.

[AVIL]

Hi,

The name is ColourLike (not to be picky, only to facilitate search).

[EpheMeroN]

I'll definately look into this ColourLike filter; especially if it can automate the color / levels matching between videos. You guys are right though. The 2 sources do not look that much different. My monitor is really old so nothing is displayed properly anymore but when I checked the images I uploaded on my laptop they looked quite similar. Do they need any gamma correction still? I've never been good with that but hear it can make a big difference in the visual quality of the video if it's set properly.

[jmac698]

It's very interesting that you mention this; I have the same problem. And my solution so far is colourlike, though I am making my own version that tries to match by Tweak() settings (hue, sat, bright, cont). It seems to me that it may be more correct to match these ways first.
My problem is different in that I want to overlay an analog, sd recording onto a digital, HD one, for example to remove excess logos, also to do tests on scaling and aspect, and compare interlacing from two streams etc.
It seems you could analyze the histogram file to find the "perfect" pattern in it, to eliminate the noisiness error, if you know what I mean. I am kind of denoising the parameters of a color change.
Would also be useful to re-remaster the star wars original back from the remastered version, or match laser to dvd versions, etc.

[EpheMeroN]

When you used ColourLike, how were the results? Did it try to find the average correct levels between the two videos? Or can you specify that you want the levels of one video to match the other, etc etc?

After looking at the sample images I uploaded again, the only thing that I can physically notice is that Video2, i.e. the XviD video, seems to be slightly brighter than Video1.

[EpheMeroN]

Will any type of color correction / gamma correction help? Hoping that someone with a better eye for spotting these types of things can help here w/ some script suggestions :-)

[AVIL]

Hi,

My two pence. I recently use Colour like:


x1=Avisource("alinea1.avi").WriteHistogram("hg1_fin.txt",100)
y2=AviSource("alinea2.avi").WriteHistogram("hg2_fin.txt",100)
y3=AviSource("alinea3.avi").WriteHistogram("hg3_fin.txt",100)

x2=y2.colourlike("hg2_fin.txt","hg1_fin.txt")
x3=y3.colourlike("hg3_fin.txt","hg1_fin.txt")



With this piece of script, clips y2 and y3 inherits luma and chroma pattern of x1 (after displaying subtracts of x1-y2 x1-y3, my eyes judged equality of luma and chroma)

After first run of the script I mark writehistogram sentences as commentary to avoid redundant generation of txt files.

Good luck

[EpheMeroN]

Would any filtering and / or a different resizer make any quality differences for the Video1 examples? I always tend to use the basic BicubicResize but I've heard EEDI2 can make upsizing look less blurry?

I can upload a sample clip if anyone would prefer that over the two sample images.

[foxyshadis]

Not going to do a thing until you're resizing at least 1.5x up or so, except slow it waaaaaaaaay down.