How to recognize types of noise in digitized video?

[colokate]

Hello! New Avisynth user here...

After spending some time learning how to write scripts with Avisynth, I came to realize that while writing scripts is a useful skill, I cannot visually identify what needs to be corrected in a video.

Noise seems to be a common issue with these old camcorder videos. So, I would like to ask more experienced Forum members about what type of noise is included in the attached video (no audio is included)? Are there chroma issues too? Is there a process to quantitively identify different types of noise or is that a matter of time and experience?

Details: This video is from the mid-1990's on a Video8 tape and was digitized in an uncompressed format using Huffyuv. The video was filmed indoors in a poorly lit restaurant which likely contributed to the noise.

Workflow: Sony HandyCam ---> Panasonic ES15 ---> Aja FS1 ---> ATI 600 USB on Win 7 (SP 1) box

Thank you,

colokate

[Feedbucket]

Mostly just experience and learning how to correlate what you're seeing with what its causes it. To my eyes there's nothing obviously bad outside of the low light noise. Often a product of cameras with smaller light sensor - less light info to record, less visual resolution and dynamic range. I think Handycams in general were limited in how good an image they could produce due to their size limitations.

Chroma flaws usually look like discoloration or coloring outside the lines. You can look at those by viewing the U/V channels individually with UToY or VToY.

Levels are in range. Maybe bump up the gamma using Levels to bring out the mid tones and then whatever spatial/temporal denoise you like (less is more, IMO).

AviSource("ExVideo_2.avi")

Crop(8, 8, -8, -16)
ConvertToYV16
#return UToY.PointResize(Width, Height)

QTGMC
before = last

Levels(0, 1.3, 255, 0, 255)
KNLMeansCL(h=2)

StackVertical(before.Histogram("Levels"), last.Histogram("Levels"))

[aramkolt]

You might try to see if you get less noise directly capturing from the S-Video output of the handycam via the ATI600 just as a test to see if other devices in your chain are contributing to the noise. Could just be that the noise there is just a side effect of the sensor not having much light to work with at the time of the recording.

You do also have quite a few stages of analog to digital and digital to analog stages here in the chain. If the handycam has a line TBC, you probably don't need the ES15 if you aren't seeing geometry issues.

[timtape]

You've correctly identified the noise (equivalent to audio hiss) just by looking. As feedbucket says it's likely due to shooting in low light. Struggling for enough light the camera turns up its gain and the electronic noise, normally masked by the well lit scene, is now revealed. This is normal.

Removing this noise is another matter. It can be reduced or removed but often at the cost of reduced wanted picture detail.
Unfortunate but true.

A well exposed Video 8 recording shouldnt normally look noisy or grainy.

[colokate]

Thank you for the feedback, Feedbucket! Yes, video quality of mid-90's Handycams was challenged I just read that Handycam sensors were either 1/4 inch or 1/6 inch. Wow. That's small.

As you suggested, I will have a play with gamma and then experiment with denoising. This may be a challenge because there are many denoising options, but it will be a good skill to learn.

And thank you for the code snippet. The UToY or VToY functions will come in handy. And I was unaware of the KNLMeansCL plugin, which aids in the removal of Gaussian noise because I understand that Gaussian noise is the noise associated with low light videos. Sometimes half the battle is knowing which tools (plugins) are available.

Cheers!

-- merged --

Aramkolt, you make a very good point that I had not considered: a potentially overly complex capture chain as I was throwing everything I had at the video. Focusing exclusively on the noise aspect of the video, it will the interesting to compare different capture chains and observe results.

Thank you!

-- merged --

Timtape, you are spot-on. I have some wonderfully exposed outdoor Video 8 videos, and they are a world away from this sample video (thank goodness). The learning experience will be experimenting with the plethora of available denoising plugins while trying to limit "damage" caused applying the plugins.

Many thanks for your reply.