Foveon® X-3™ Color Image Sensor Technology

Getting Clear Color

In monochrome image sensors, each pixel generates a signal representative of the light arriving there.  Thus, the sensor output is simply a sequence of individual pixel measurements of intensity easily assembled into a matrix for analysis or display. 

Color imaging has, until now, been more spatially and optically complex than monochrome imaging requiring some type of applied filter to partition the incoming light into three bands and then carefully controlled alignment to preserve or intensive calculation to reconstruct the color data at each pixel.

In the least complex color cameras, a single monochrome sensor is sequentially covered by three color filters, producing data that is accurately registered but separated in time.  Such cameras cannot image motion and use light inefficiently.

To maintain the pixel alignment but allow for motion, three monochrome sensors can be mounted on a color separation prism.  This arrangement also makes good use of the light because the separation filters only absorb a small fraction.  Prism assemblies, though, are expensive to build because they require three sensors and precise alignment.  Their long optical path length also limits the selection of optics.

With its three stacked layers of photodiodes, the Foveon X3 technology alone combines the optical simplicity of monochrome imaging, the geometric accuracy of sequential color and the efficiency of the prism assembly in a single CMOS device.  

 Matrix Image Sensors

The majority of color cameras now use color matrix sensors because they are inexpensive and simple to use with most optics.  In these sensors, each pixel has an individual color filter, most commonly in the Bayer pattern shown in this illustration.

Color filter arrays lead to two important problems.  First, the filters absorb most of the light, passing only that part desired to sense a specific color band.  In addition, since each color band is detected in a geometrically separate location, reconstruction of the color for each pixel requires estimations and assumptions that can only approximate the actual color information received at the center of the pixel group.  The resulting color errors appear as color aliasing that is impossible to remove.  Aliasing can be reduced by blurring the incoming image but only at the expense of overall sharpness.  Finally, because reconstructing the image requires calculating the two missing color values for each pixel, either the offset in the data must be ignored or a very large number of computations must be carried out.  Neither approach accurately represents the incoming image.

Foveon X3 Image Sensor

In the Foveon X3 image sensor, three photodiodes are formed in every pixel, stacked like the three layers in color film.  This arrangement utilizes the wavelength-dependent light absorption property of silicon to produce natural filters that use the incoming light to greatest advantage.

When every pixel senses all the color, artifacts from color filter offsets are eliminated.  In the illustration, the color banding is gone but a close examination shows that aliasing just like that from an unfiltered monochrome sensor is still present.  Because no optical blurring is needed to control color artifacts the full resolution of the sensor is always available.  The color resolution of Foveon X3 image sensors is identical to their monochrome resolution so there is no need to reconstruct missing color data by complex computation.

Download the complete X3 brochure 

See more images

All images are copyright Foveon, 1998-2002. Permission must be obtained prior to use.

Files marked require Adobe Acrobat Reader.

High-Tech Digital has been delivering Imaging and Machine Vision products since 1983. To find more about our imaging systems and components call 310-265-8203 or send an e-mail to:

info@high-techdigital.com