View Single Post
  #2  
Old August 25th 15, 06:13 AM posted to rec.photo.digital
Sandman
external usenet poster
 
Posts: 5,467
Default New sensor with unlimited dynamic range

In article , Alfred Molon wrote:

http://web.media.mit.edu/~hangzhao/modulo.html


When a pixel fills up, it is automatically emptied. The camera
counts the number of times a pixel is emptied. Pixels in the darker
image areas probably never completely fill up.


Pretty smart, but the example images didn't really sell the concept enough. With
"unlimited" dynamic range, you'd think nothing in the image could possibly be
overexposed, but some of the example images, while a lot better than the normal
photos, still looked like they were blown in areas.

Since this "unlimited" dynamic range only works for highlights, you also need to
expose for the darkest part of the scene.

Or, if this technology would find itself into normal Nikons and Canons, they
would automatically expose for the darkest part of the scene, while everything
else would be blown and then in-camera recovered.

Not sure what the coloring is used for in the example images? I am assuming we're
looking at raw images with a really high bit rate, and for some reason they are
showing those higher values with funky colors for some reason. Makes me wonder
what kind of post-processing is needed for each photo.

The logical thing to do is to use a high bit depth for the image, like 16 or 32
bit, but keep "normal" exposure data in the first 14 bits for example. So instead
of 0-2744 in 14bit be the same range as 0-4096 in 16 bit, the first 0-2744 in 16
bit corresponds to 1 bit data, while the rest is used for the otherwise
overexposed data, which of course requires some fancy post-processing to even the
range out to a normal image.

Or you use a 16 bit readout from the sensor (as per above), then in-camera re-
arrange that to a 14 bit range, producing a balanced image.

--
Sandman