View Single Post
Old February 5th, 2008, 09:34 AM
Nikolai Sklobovsky Nikolai Sklobovsky is offline
Senior Member
Join Date: Aug 2006
Location: Thousand Oaks, CA, USA
Posts: 1,181

Originally Posted by Asher Kelman View Post

Hello reality! Of course this is the correct strategy in principal but the perhaps wrong path. CMOS sensors can be made so that each pixel is independently controlled like a single cell camera or binned to be a multi-cell camera. Thus one can throw out ISO concepts except for the pixels receiving less light. Pixels can easily be shut off in at the rate of 1/150,000 sec. Other pixels can be collecting photons as long as possible.

Companies like Texas instruments are likely to be putting such ideas into mass produced $300 cameras since the market scale is so massive. Grandma with her little digicam will take pictures that span bright sky and water to a toddler in the shade of a beach chair.


Pixel-based ISO control is a very cool concept, but it also implies that the camera should recongnize the goal of the particular image. Grandma on the beach example is fine, but imagine a studio shoot going for the highkey or the lowkey look. In the first case camera would try to turn the blown bg into nicely look light gray (220,220,220), showing all the possible wrinkle of the not-so-perfect backdrop, and in the second one it will try to boost the bg into something like (20,20,20), again showing all the possible bg defects and defeating the very goal of the image. I think this level of the AI is still eons ahead of us.

But Panasonic the idea instant bracketing, ie.e of having several images from one shot - me likie!

I think both ideas have merits: pixel-based correction can be used in consumer P&S while the instant bracketing could improve the life of the pro/prosumer world.
Sr.Software Engineer
PhotoSoCal, Digital Grin, LA Shooters, NAPP, PPA
Reply With Quote