LINLOG: global shutter and highlight compression

[lordtangent]

LOG sensors have long been used in machine vision, but until now they were not too practical for video due to the long integration time which caused severe rolling shutter artifacts. The LINLOG technology mentioned in the following link changes all of that:

http://www.photonfocus.com/html/eng/cmos/linlog.php

A digital camera using LINLOG could actually exceed the dynamic range of film. No hard clipping. No "black sun". Just smooth detail all the way in to the highlights and a nice film-like roll-off (AKA shoulder).

The advantage also works out to more efficient use of code space in the RAW files. Basically, they have a log-like compression that saves bits. And since the log compression happens in analog space (before sampling) it requires no extra circuitry outside of the image sensor to achieve.

Comments?

 

[Damon Meledones]

Looks pretty awesome to me!

 

[Jason Diamond]

that look amazing. wonder if Graeme has any thoughts?

 

[Graeme Nattress]

My thought is I first saw that page a few years ago....

Graeme

 

[Jason Diamond]

exactly why i asked for your thoughts!
so its not a feasible tech or just isnt practical?

 

[lordtangent]

Yes, I want to hear from someone who makes consumer products why this technology hasn't found it's way into any (main-stream) consumer products even though it is considered "old news".

What is wrong with it?

 

[Michel Hafner]

My thought is I first saw that page a few years ago....

Graeme

-> RED has purchased the technology and uses it in Mostro, :w00t: :biggrin:

 

[lordtangent]

Yeah...

Seriously, why isn't this being considered? It's like the silver bullet for fixing what's wrong with digital.

 

[lordtangent]

The fact no one from Red has anything to say other than this is "old news" really makes me wonder what is up.

 

[Mark Crabtree]

Given that, unlike CCD's, CMOS sensors have a seperate amplifier for each pixel, is it theoretically possible to ramp down individual pixels sensitivity after they achieve 90 percent exposure in ten log steps at 91 to 100 percent? Each step would get ten times less exposure resulting in a shoulder effect.

 

[lordtangent]

exactly. So... I'm wondering why the only feedback we get from anyone affiliated with Red is sort of obtuse.

I want to think that's what they are doing... But it would be wild speculation to simply assume that since they don't publicly acknowledge something that seems like a good idea, it is exactly what they are working on.