h a l f b a k e r yOK, we're here. Now what?
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
A round 360 degrees camera that using image processing
follows and focuses on only a certain area, just like the eye
goes jumping to various regions of the images darting back
and forth in Raped Eye Movements scanning whatever it sees,
discarding most of the info, and focusing only on the
important
parts.
Why can't a camera do the same? - But with no mechanical
parts!
The end of the scanner, the unfocused images and all other
Photographer Problems. Welcome to the Camera Problems
era.
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Annotation:
|
|
"Raped Eye Movements" sounds somewhat discomforting... |
|
|
// the eye goes jumping to various regions of the images
darting back and forth in Rap(i)d Eye Movements scanning
whatever it sees, discarding most of the info, and focusing
only on the important parts // |
|
|
I hate to sound like a know-it-all, but this is called
'sacceding' (sp? Root: 'saccede', pronounced like 'succeed'),
and it is one of the things that robotics engineers have
struggled to replicate for many years. Apparently it is a
major stumbling block in allowing an advanced
robotic eye to gather and process visual information as
quickly and effectively as we do. Or so I have read, at any
rate. |
|
|
R.E.M. is something different; depending on who you ask,
it's thought to be either a reflexive or somatic behavior
linked to the perceived visual stimuli in our dreams, or
simply the random firing of neurons that takes place in
many of our motor centers while we rest (like twitching
fingers, shrugging, etc.). |
|
|
//follows and focuses on only a certain area// How? If you are not going to have any moving parts, then the sensor of the camera must be able to see the entire field of view at the same time, in focus. |
|
|
Now say you get that far. How does the camera know what is of interest before it does any image processing? It happens that the human brain does preliminary processing on the out of focus/out of interest area and uses that to indicate the next point of interest (plus lots of rapid random movements around the point of interest to fill in data). In order to develop that sort of data on a camera you are still going to have to process the the entire image. |
|
|
//I hate to sound like a know-it-all, but this is
called 'sacceding' // You don't - it's "saccading". |
|
|
You could sort of get there by just having a
camera that captured a much wider field (a fish-
eye lens) and then saved only the interesting
region. But, as [MechE] points out, how does the
camera know which is the interesting region?
Plus, for any given level of sensor technology,
you'll have fewer pixels in your region of interest
than you could otherwise have. |
|
|
However, I guess that as pixel-count gets higher
and higher, there'll come a point where it makes
more sense to take a wider image than to increase
resolution. |
|
|
(But, before we expend pixels on wider views,
they'll be used to get greater depth of field -
Google "light field camera", which uses extra
pixels to capture light from different directions.) |
|
|
Sorry, drifting into a ramble here, but... |
|
|
the fundamental problem is that we expect
cameras to be as good our eyes, but our eyes
depend on really crap optics and very very clever
processing. You never see what you're looking at:
what you see is the model your brain makes,
based on jumpy, patchy, blurry images and a lot of
filling in. That's very hard to implement in a
camera. |
|
|
Let the camera saccade to anything that moves,
and
let the computer generate an image in which
anything not moving is assumed to look the same
as
last time you foveated it*. Wouldn't work in all
environments, but in some, at least, it might do a
credible imitation of primate visual system's
\\model-building with crap optics\\ |
|
|
*Don't video compression schemes work that way?
Essentially, this system would record "pre-
compressed" video. |
|
|
Thanks, [Max], I knew I had it wrong. It's just that I've
spent a long time cultivating my largely undeserved
reputation for never voluntarily fact-checking or citing my
sources, and I'm not about to spoil that over a simple
mispelling when I'm busy holding forth on a topic others
know far more about. |
|
| |