Faking and expensive tilt-shift lens

["Followup-To:" header set to rec.photo.digital.slr-systems.]
Paul Furman wrote:

I'm not sure how a half pixel shift would work for increasing
resolution... perhaps you'd need to go 1-1/2 or 2-1/2 pixels.

You may be interested in
http://auricle.dyndns.org/ALE/gallery-auto/
http://auricle.dyndns.org/ALE/download/ale-0.7.x-tech/
http://www.wisdom.weizmann.ac.il/~ir...SR_CVGIP91.pdf


-Wolfgang

On Tue, 10 Feb 2009 21:49:15 +0100, Wolfgang Weisselberg
wrote:

John A wrote:

G R
B G

Together, those four photosensors make up one pixel.

Sorry, your claim is as invalid as your email address.

Yeah, I figured that out. :P Read on.

"mcdonaldREMOVE TO ACTUALLY REACH wrote in message
...
| Paul Furman wrote:
|
| Dissolving the filter doesn't change the circuitry and firmware. The
| camera still interprets the 2x2 sensor array as one pixel.
|
| That's incorrect. The algorithms used to determine pixel values
| actually look at MORE than a 2x2 array! They try to determine, for
example,
| if an edge runs through a pixel. If it does, they use other
| parts of the two sides ... outside a 2x2 array ... to try to determine
| all three RGB values of the pixel. The algortithms are nonlinear ..
| they do not merely interpolate. They actually GUESS.
|
|
| The raw file has the separate bayer parts unmerged.
|
| I'm not sure how a half pixel shift would work for increasing
| resolution... perhaps you'd need to go 1-1/2 or 2-1/2 pixels.
|
|
| 1/2 pixel increments would work.
|
| Doug McDonald

and the nyquist limit?

"k" wrote in message
...
and the nyquist limit?

Increases directly with increased bandwidth.

MrT.

"Mr.T" MrT@home wrote in message
u...
|
| "k" wrote in message
| ...


again

1/2 pixel increments would work.



| and the nyquist limit?
|
| Increases directly with increased bandwidth.
|
| MrT.
|
|