h a l f b a k e r yI think this would be a great thing to not do.
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,
|
|
|
Some time ago I used some software to stitch together a panorama from multiple views of a scene. One of the features of that software which impressed me was how it handled the overlap between pictures - it basically worked out the distortion on each frame and corrected it so that the overlap was contiguous
with the central area from the frames on either side, i.e. it removed the edge distortion.
Another piece of software I have used takes two images of the same scene, and superimposes them intelligently, giving a better resolution picture than the original.
I realised this evening that similar algorithms could be used to reduce the distortion and improve the quality of images taken with cheap digital cameras (i.e in the under $50 range rather than the > $500 range such as the Canon Rebel XT)
These small & cheap cameras have almost all the features of their big brothers - except a large glass lens.
My theory is that if we have a camera with four of these cheap lenses (arranged in a 2x2 square), and four CCD pickups, we can stitch the four images together with software such as already exists and create an image with quality much closer to that of a camera with a large glass lens.
The reason for doing this is primarily size, not cost or quality. A large lens will always be better (unless you have so many small lenses that they combine to be equal to one large one) but you can't fit a Rebel XT or an EOS in your pocket - with one of these, you could.
(This is not the same idea as the previously posted "swarm camera" or any of the multiple-lens cameras whose reason is to have a different exposure time on each lens, nor is it a 3D camera. Note that four pickups with separate lenses are not the same as one pickup with four times the pixels; it may be in terms of resolution but not in terms of image distortion from the use of small fixed lenses)
Graham
this one takes 9 Pics at same time.
http://www.opticspl...to-camera-pop9.html [jhomrighaus, Sep 19 2006]
The BIG version of your camera
http://graphics.sta...edu/projects/array/ [jhomrighaus, Sep 19 2006]
A patent for same.
http://www.freepate...ne.com/6954310.html [jhomrighaus, Sep 19 2006]
Compound Eye Camera
Compound_20Eye_20Camera [xaviergisz, Sep 19 2006]
Autostitch
http://www.cs.ubc.c...orama/panorama.html Multiple-image stitching [gtoal, Sep 19 2006]
Image stitching
http://www.hpl.hp.c...ama/SPIE5242-31.pdf More detail on stitching in this paper (PDF) [gtoal, Sep 19 2006]
Improving resolution by image registration
http://clluengo.lbl...%20Registration.%22 sub-pixel sampling [gtoal, Sep 19 2006]
speckle imaging technique (fast sequential frames)
http://www.llnl.gov...June05/Carrano.html Tangentially related, posted here as an interesting read [gtoal, Sep 19 2006]
Superresolution
http://lcavwww.epfl...re/superresolution/ free image-enhancement software for MATLAB [gtoal, Sep 19 2006]
pretty pictures
http://www.astrosur...m/cidadao/super.htm Readable paper on astronomical image enhancement by subpixel registration [gtoal, Sep 19 2006]
High dynamic range photo gallery
http://www.cambridg...mbridge-gallery.htm [ldischler, Sep 19 2006]
SupaImage2
http://www.imageip.com/biz/SupaImage2/ commercial software to improve resolution by superimposing two low-res images (eg from a webcam) [gtoal, Sep 19 2006]
[link]
|
|
On the surface this seems like an outstanding idea. Would your camera require four times the available memory, however, in order to be able to accomodate the "stitching" and dithering? And, if so, doesn't this get you back to the price point of the more expensive digital SLR? |
|
|
[ And, if so, doesn't this get you back to the price point of the more expensive digital SLR? ] |
|
|
If the post-processing could be done in a custom chip, then you would only save 1/4 of the data that went into taking the picture. All of these digital cameras now use a COACH (Camera On A CHip) so positing a custom chip isn't as expensive as it sounds. |
|
|
*Or*... the four CCDS could be half size, and the synthesized image be full size, so the storage requirements would be the same - in which case you could do the post-processing on your PC rather than in the camera (sort of like a 'digital developing' stage) thus avoiding the requirement for the custom chip (especially if it turns out the processing requirements are hefty and too slow for interactive use) |
|
|
*Or* you could lash 2 cheap cardboard box cameras together with double sticky carpet tape, have one with a blue filter and one with a red filter, take pictures simultaneously (using a special fork to press both buttons at once), then synthesize the two images by taking a third photo (with a different camera) using a reversed binoculars and 3D movie glasses. One would wind up with a perfect 3D digital image. You would have to add the colors back using the Paint program. |
|
| |