Generate test pattern from video

[Mounir]

The more i attempt to fix videos the more i am convinced that without test patterns (http://en.wikipedia.org/wiki/File:SMPTE_Color_Bars.svg) you can't do good work (i mean excellent work / pro-like work).
Colors, luma this is math ( U-luminance, V-luminance etc... ) the gurus here know better than me and with the proper tools (histograms,video scopes (http://www.youtube.com/watch?v=R6jO3ui_FbY)) you can do wonder assuming you understand what they show.

Now onto the subject, i have many commercial videos but not all of them have test patterns so i'm wondering if there is a technique available to, somehow, generate them by analysis/averaging the results from a given video and have something useful to work on.
If there isn't one i suggest someone investigate this possibility in great depth

I'm all ears gurus.

[Ghitulescu]

I'm not sure that I understand your question. Test patterns are used to calibrate the tools (player, monitor) not the videos.
Are you referring to histograms (ie spectral/spatial distribution)? Pro/semipro tools (premiere and up) have also vectorscopes as modules or plugins. But these are not to calibrate the video but to make sure you won't output a non-standard video.

[Mounir]

I understand your argument but in my opinion they also help. Ideally one would work frame by frame, adjust whetever need a fix but it if you have 30 thousands frames it's another story. I suppose i'm asking the moon.

[poisondeathray]

They don't help in the context you're referring to

You use color bars and color charts in the beginning when shooting the video, so you can match different cameras (A, B ,C etc...)

A generated color bar after the recording is independent of the data in the already recorded video . There is no relationship.

Vectorscope, histogram, waveform monitor are useful in color correction and grading, but this is not the same thing as a generated test pattern

[Ghitulescu]

If I understood correctly the question you didn't ask, I think you want to achieve what the premiere plugins Broadcast Colors or Video Limiter do. Both of them modify the video to be standard-conform, something like not too much white, no darker than black neither higher saturation.

[Ghitulescu]

http://forum.doom9.org/showthread.php?t=160503

[jmac698]

You an analyze a video for white balance or to fix out of range levels. When using capture cards, I've analyzed video to find "missing codes" which would be a sign of a non-linear preamp or perhaps bad ADC. What I mean is I'm doing a histogram. You can check a video for clipping, meaning white is at some other value than the standard 235, in which case you could safely increase contrast. You could check for specific problems like a rotation of chroma which might happen in analog circuitry, with composite or svideo connections.
But generally I can't think of a way to apply your idea. Good thought though!

[Mounir]

I didn't notice untill now there is a script you've made actually but i'm affraid i don't understand how that work there is very little instruction but i suppose you have to burn the test patterns on dvd then copy from dvd player to a vcr problem is my dvd player is pal so it'll send either pal or pal60 i'm affraid which will be counter productive.
If you could make the same test with an smpte 170m i'd be more interested

[jmac698]

Glitchanalyzer is at
http://forum.doom9.org/showthread.php?t=84481&page=5
And the instructions say:

#Use:
#1. Modify this script to set mode to generate and type to "NTSC" or "PAL" then encode the resulting diagnostic video
# If encoding to mpeg2 use interlaced, top field first (e.g. in HCencoder)
#2. Capture the diagnostic video (e.g. in VirtualDub)
#3. Set mode to analyze, set the captured video filename and the timecode log output filename
#4. Run the analysis by playing back this script once (i.e., load the modified glitchanalyzer.avs into MediaPlayer and play)
#Load the resulting timecode.txt into a spreadsheet and
# note where the timecode skips or duplicates numbers (a perfect capture will count from 0 to 15 repeatedly). It is up to you to interpret this.

What part was unclear, I'd like to be a better writer!
My script works with PAL. It can still detect missing, reversed fields, or dupped frames in PAL.
And yes, just set the mode to generate, type to PAL, make the video file, record to VCR, capture the result, set mode to analyze and set the filename, then read the output. Maybe I missed the part about recording to vcr?
Anyhow, by SMPTE170 do you mean just colorbars() which is also called Rec601? That's easy to do, I'm sure I have recordings of colorbars from a VCR around somewhere. A program Dscaler has a builtin calibration tool, based on playing certain test disks like brand Avia or Digital Video Essentials (but work with normal colorbars()).

You mentioned VCR but, to find capture problems in the capture card you can still record just from the DVD player. I just use a mediabox, it's quick and simple to plug in a USB with my test pattern as avi. Even with the stable timing of the mediabox player, the capture card still has problems.

You can do both at once; I can't remember now what I use for a background but you can use colorbars in the background too, the glitchanalyzer stuff is overlaid on top of it.