Open letter to a video cam hardware manufacturer

#1

I’ve read your rules, went for [Technical and Hardware], but then saw [Off topic], which is probably best suited for this concept/idea. I will accept your ruling of what I consider is best, this approach, which may have your interest.

Open letter to video hardware manufacturer.

Dear company like Panasonic,

Could you develop a video camera with a lens’ focal length going back and forth? That is, focus between infinity and macro (close up), constantly going back and forth capturing all distances.

You could link corresponding frames*, and compute depth-lines (based on focus and sideway shifts distances), right on the camera. *#focus

Perhaps consider developing a YouTube required 5-in-1 technique, video based on highest (submitted) resolution, if ever YouTube would accept (already) optimized material. Just fine if the material should have other anchors in it; al long as it’s accessable to ‘you and me’, everybody in the community. Here 3D anchors (grips/tags) are very welcome. (For example for linking, attaching or super imposing. Otherwise video material would be just many frames of paintings.

(Perhaps your cameras can detect lines too, for example, recording vector on top of bitmaps, which are normally lost in the footage, but for now that is taking things a little to far, though nice to think about.)

Please take a look at it and see whatever is yours.

Bas $

Definitively not a CG expert, but definitively with care.

#2

The Lytro Lightfield camera does this:

https://vimeo.com/161949709

More info here:

www.lytro.com

#3

Thank you for your post skeebertus for letting me know, and thanks developers for this far more advanced and realized technology and tools than what I had in mind. Great! For me it’s inspiring.

This (Lightfield Solution) camera shouldn’t come cheap, could be a life saver anyway so to speak, so perhaps budget camera’s can still drive their focus back and forth as fast as possible, but, this Lytro Cinema camera is absolutely fantastic. I’m a bit unprofessional right now.

Very much appreciated, as is all the envelope pushing creative artwork on this website, if I can mention this just once. Great grip…

#4

A bit more theory which may prove to be false, you’ll probably see that soon enough, then I apologize for it, or an alternative to unblur (focus) existing footage as much there’s information to be calculated (not too much compression ideally), potentially making it a bit 3D as well:

Existing (old) footage detail can be accumulated where the picture is out of focus; everything that’s out of focus is blended from many colours to less or one, and slightly different, shifted, in the second frame. That difference in colour can be subtracted, to figure out the difference where it’s overlapping, or just mark the area with the new colour, getting a smaller area every time because more colours are discovered. (As is pixelated parts of a video, I think they should add something in that does not correspond to the original, like the black bar which has no relation to what it covers up but protecting a suspect or accused. You’ll see that familiar pixelated footage rapidly taking different blended colours from whatever it covers behind it.)

Well, if it’s true, security cannot be achieved by obscurity, also in this case, or (by) darkness for that matter. Supplementary to Autodesk’s 123D Catch where candidate objects changes shape in other frames to be matched (?, or perhaps it’s cloud-point based, I didn’t describe that too well, but) just to say that’s different to what I mean.

But the footage has to be of good quality, and not ‘shot’ or gone bad. It’s still material with which to work with.

#5

What you are talking about is usually called Super Resolution.

Like this software for law enforcement:

https://www.motiondsp.com/ikena-forensic/

also

https://www.motiondsp.com/ikena-isr/

There are a number of techniques for this:

  1. Sharpening/Deblurring/Deconvolution algorithms that sample a 16 x 16 pixel area or larger (in your LED TV for example)

  2. Algorithms that combine image details found in multiple similar video frames or images

  3. AI algorithms that “guess” or “generate” detail that MIGHT be in the blurry area

Look here:

http://decsai.ugr.es/pi/superresolution/superresolutionv1_large.png

http://photoacute.com/studio/

http://download.cnet.com/QE-SuperResolution/3000-2169_4-10392416.html

https://www.motiondsp.com/ikena-forensic/

Or were you asking about a different method?

#6

Thank you skeebertus,

sorry for my late response; I did ‘(re)visit the forum’ but wasn’t notified since, so I assumed this treat went quiet. Good to see that’s not the case. Although I would respect it.

Thanks for your links and accompanied text; I’ll go through them first. I think it’s very interesting, already existing or not, and happy to share with creative minds and general usefulness, if I’m not mistaken.

Thanks in advance for your structured, to the point, post. I’ll try to match.

#7

The image enhancement is exactly what I mean, and software that reveal pixels/detail/colours that might be there helps definitively; the main thing is (that) I think pixelated video can be engineered back/reversed, and out of focus parts as well, depending on the quality of the video, the compression artifacts, and the amount of blur/out of focus. The more out of focus, the less I think is recorded in blends of colours, very precise/exact when pixelated.

It looks like Forensic Video Enhancement Software from Ikena does that. Thanks for letting me know about this software.

I have no thoughts how this can be done on just one frame/image. I need more time though to check all your links fully, and thanks.