Klaas Decanniere on wrote...
|
|
| Hello,
|
|
| I am trying to make night shots with a ccd camera.
| The maximum exposure time that can be set on the camera is 750 ms, which
| is of course too short.
|
| Assuming the noise is random and the object I want to see does not move,
| I should be able to improve signal-to-noise by taking the same shot
| several times and adding all the images together (using an iterative
| shell script which adds each successive image to the "sum" image,
| starting from a black image with the correct size and depth): the noise
| should "cancel out" and the signal should accumulate.
| As adding many layers of noise will eventually saturate the image, I
| decided to use the quantum:format=floating-point option.(support is
| compiled in) and I save to the "miff" format.
|
The operation is -average
convert image1.jpg image2.jpg image3.jpg -average merged_image.png
However this will only work if nothing has moved. Not the object, the
background, the camera or anything.
If something has moved then you may need to use the new -distort
operator to distort two of the pictures to align it exactly with the
other picture.
Multiple control points can be used for most distort operations (like
perspective).
However at this time the distort operator is needing some extra work by
me to cleanup the underlying image filters. The default filter it is
using is 'gaussian' which is very blurry, and while other filters are
implemented, they are not fully tested or verified in correct working.
This is my next project in the Magick Core, along with improved handling
of 'infinite' or horizon views, as well as 'arc' or "othogonal to polar"
conversion.
I also want to rework the bilinear polynomial distortion, and add a more
general higher order polynomial distortion, and posibily add 'grid',
'mesh', and other local distortion type methods that can be used in
future higher order operators implementing for both spatial and color
image "morphing" such as use in all the time in movies.
| Now for the questions:
| - is it correct that both the "display" and the "identify" programs and
| the "histogram" option of convert have trouble handling this
| floating-point format?
Histogram has not bee checked with floating point as far as I know,
but should still work reasonably well. It is in fact something that
also needs some work, such as the addition of 'cumulitive histograms'.
The other programs "display" and "identify" should work but may need
user testing.
| - supposing the adding worked fine, what would be the best way to
| convert the result to a "normal" image for viewing on the screen (8 bits
| per channel) while preserving as much information/contrast/image quality
| as possible (without a priori knowledge about intensity distributions or
| object shape etc.)
|
That is the problme that IM was initially created for!!!
| P.S.: I correct each image with bias and dark images before adding them
| to the sum
Probably a good idea. You may like to look at Fred Wienhaus's color
improvement filter programs.
Anthony Thyssen ( System Programmer ) <[EMAIL PROTECTED]>
-----------------------------------------------------------------------------
"Violence never solves anything." --- Attila the Hun.
-----------------------------------------------------------------------------
Anthony's Home is his Castle http://www.cit.gu.edu.au/~anthony/
_______________________________________________
Magick-users mailing list
[email protected]
http://studio.imagemagick.org/mailman/listinfo/magick-users
