Inspired by the single pixel digital camera (see link). Instead of focusing the light on a regular single pixel color sensor, you could put the incoming light through a rotating prism that separates the spectrum of the incoming light and measure the strength everywhere in the spectrum.
The resulting photo would tell you so much more about the nature of what you've taken a photo of then a regular color shot. To an extent, you would know the chemical makeup of what the photo was taken of, especially if the subject was illuminated by broad spectrum black body radiation.
Another thing you could add is a rotating polariod filter, to determine polarization. This might tell you something about reflections in your photos.-- jmvw, Nov 01 2006 Single pixel digital camera http://www.physorg.com/news79019816.html [jmvw, Nov 01 2006] So is this a single pixel spectral camera or a regular CCD array spectral camera?
You could use a set of parallel, rotatable, elongate prisms over the top of the CCD array to separate the spectrum to each pixel.
You may not be able to do proper spectral analysis because of attenuation of various frequencies of the lens system.-- xaviergisz, Nov 03 2006 I don't think anything as complex as a rotating prism is required. Most digital cameras already interpolate the image. An 8 megapixel camera will typically have 4 million red (er wait... no... green) pixels, 2 million red and 2 million blue. The camera's firmware will spit out 8 million RGB pixels.
If you want to spinning prisms and twirling polarisers, all you need to do is increase the diversity of the pixels -- that one is broad spectrum red with 60° polarisation. That one is narrow blue around the colbalt line with no polarising filter.
[+] because I would to have a camera that sees things I can't. (as xaviergisz says, though, current lenses are designed to correctly focus and transmit only visible light.)-- st3f, Nov 03 2006 //So is this a single pixel spectral camera or a regular CCD array spectral camera? //
It's a single pixel spectral camera. The singular location of the sensor should make it easier to build accurate optics for wavelength selection.
I think NASA would have liked this camera on the Mars rovers, instead of those filters.-- jmvw, Nov 03 2006 Those are terrific photos, Ian.-- jmvw, Nov 03 2006 What would be the equivalent of a flash?
[st3f]//Most digital cameras already interpolate the image// So, extrapolate the image from the blurred colour from a single pixel by deconvolution?-- Dub, Nov 06 2006 //What would be the equivalent of a flash?//
You'd want a true full spectrum flash. I believe black body radiation is broad spectrum and a thermal flash such as a magnesium flash might do well. Incandescent light would work well too, if the filament is hot enough to get into the ultraviolet. The old single use flash bulbs might work great. The camera could compensate for a somewhat uneven spectrum of these lights. A regular xenon flash could be used, but the spectrum would be limited to the lines of the gas and you would end up with little more then a good old color photo.
//as xaviergisz says, though, current lenses are designed to correctly focus and transmit only visible light.//
Regular camera's do excellently in near infrared which is commonly used for spectrophotometry. Digital camera's have a filter glued to the CCD to filter out IR. Some people replace this filter with a piece of dark (exposed and developed) color negative film to make digital IR camera's. Considering that UV filters are common, I think camera optics may work in near UV too.
I doubt that focus of a quality lens is very dependent on wavelength. If it were you would have some color shift even in visible light. Ian's crisp IR photos also illustrate the same focusing in in visible and IR. When he took these photos, he focused using visible light.
Note that this camera wouldn't necessarily range to IR and UV. The important part is that it records a detailed spectrum rather then some broad ranges.-- jmvw, Nov 06 2006 Dub: I was asssuming that the camera would have more than one pixel. I'm bewildered why the idea is restricted to that when so much more is possible.-- st3f, Nov 06 2006 st3f, please follow the link about the single pixel digital camera. It's a recently developed, different way of producing a full image.
As I understand it, it uses a randomly oriented array of tiny mirrors to reflect light from a variety of angles onto a single pixel light sensor. It takes a number of samples with the mirrors re-oriented randomly each time and software reconstructs the only image that would have produced this combination of samples.-- jmvw, Nov 06 2006 I was hoping this might be good for photos of ghosts.-- bungston, Nov 06 2006 You couldn't use a flash for a spectral camera since exposure time would be at least a few seconds to do a spectral sweep.-- xaviergisz, Nov 06 2006 Good point. :)-- jmvw, Nov 06 2006 [jmvw] I love the idea of a black flash! :) Put on the kettle- Fill the cube - I'm going to take a picture.
st3f: I was half-baking... Wondering, if the blur is made up of convolved image data, why you couldn't deconvolve it to re-create the scene - Using 1 pixel's cool, because, presumably, there's no need to be so careful about the RGB quality of that sigle pixel (when you have a vast array of 'em, it's quite important that they all behave the (roughly) same, I guess) - As all the pixels would be re-generated from the same (possiblly, RGB-wise, skewed) pixel, and that could be corrected after the image had been re-created... BTW most of this might be b0llux :)-- Dub, Nov 08 2006 I like it! Even though I thought, when I clicked on it, you were suggesting taking pictures of ghosts. That'd be cool, too. But I think your system would be hard to make, and it would be hard for the average viewer to know what the heck you picture is. NASA WOULD love it!! I like my normal camera, though. It's simple.-- TahuNuva, Nov 02 2007 [+] I want a DC to X-ray camera - It should compress all the frequencies (non-linearly, throughout the whole spectrum) to just those we can usually see. What would I do with it? Dunno, yet - But I want one.-- Dub, Nov 03 2007 Hyperspectral imaging is baked, WKTE, and used in space for decades. A hyperspectral camera that captures the whole cube (width×height×wavelength) at once was even proposed in 1938, according to Wikipedia.
I guess the novelty here is just to make a hyperspectral camera using the DMD coded-aperture technique (which replaces the spatial sensor used in most cameras with a global sensor and a spatial modulator). That would make some aspects of building a hyperspectral camera easier for less-sophisticated builders (such as amateurs) versus other types of hyperspectral camera, so [+].
I happen to have a DMD with some stuck mirrors. Maybe I'll see if I can get that working someday.-- notexactly, Dec 18 2018 I once built a coded aperture spectral camera, with a mask I made from orthographic film (took some pictures of aperture patterns, then developed it, used the negative as a mask). It didn't work that well, the silver wasn't high enough contrast.
The reconstruction was done in matlab, but I suppose a modern smartphone would be able to do it. I'm not convinced there is a market for consumer spectral imaging though.-- mitxela, Dec 18 2018 random, halfbakery