In the D20 demonstrator we use a CMOS chip with a Bayer mask, which is
one of the many forms of Color Filter Arrays (CFA) that have been
devised to allow three colors to be recorded with a single chip. On a
Bayer mask chip each pixel has one color filter, one next to the other.
The first row of pixels has green and red color filters (green, red,
green, red, green, red, green, etc. ), the next row has blue and green
filters (blue, green, blue, green, blue green, etc. ) and the next one
again green and red, and so on.
row 1: G R G R G R G R G R ...
row 2: B G B G B G B G B G ...
row 3: G R G R G R G R G R ...
row 4: B G B G B G B G B G ...
...
The raw "Bayer data" is essentially a monochrome
image where each pixel corresponds to only one specific color value. In
order to get a color image, the colors have to be "reconstructed" based
on the Bayer data.
f you look at the pixels you will notice that each red pixel, for instance, is
surrounded by four green and four blue pixels. Also, because there is an overlap
in the color spectra of red, green and blue, the available red value is at least
in part the result of light in another color. Based on the knowledge of what the
colors and values of those neighbor-pixels are, and based on the knowledge of
the overlap in the color spectra, it is now possible to work out (reconstruct)
what the green and blue values for that red pixel should be.
This process is more accurate than the
interpolation used to increase the size (i.e. pixel count) of an image. In
interpolation, completely new pixels are "made up" based on what the
neighboring pixels look like. In Bayer data reconstruction we already have
pixels, we just don't know two of the three color values. Since we do know
the colors and values of neighbor pixels and since there is a color spectrum
overlap, we can reconstruct the missing information very accurately.
Please note that the actual color reconstruction is more
complicated than the method described here. For instance, to determine a
given color value for a given pixel we use more than just the eight
neighboring pixels. Furthermore, it is also possible to improve the result
by incorporating certain assumptions about real world images in the
algorithms (e.g. colors coincide at edges, etc.). We have simplified the
process in this description to aid in understanding.
But what about resolution? First of all, we have to clearly
define resolution. Resolution tells us how small the smallest structures
(e.g. alternating black and white stripes) are that an optical or opto-electronic
system is capable of reproducing. In digital photography, there is a
tendency to describe resolution in terms of the number of pixels on the
chip. Depending on the technology used however, the actual pixel count of a
chip does not directly correspond to the resolution the system is capable of
reproducing. The D20, for instance, is designed to accurately reproduce
images at HD resolution (1920 horizontal pixels). In order to achieve this
goal, a Bayer mask CMOS chip of a higher pixel count is necessary.
On the Bayer mask chip itself the full number of pixels is
not available for each color. For a 2880 x 2160 chip, the red channel for
instance does not have a resolution corresponding to 2880 x 2160 pixels. One
could assume that since every second pixel is red in every second row, we
have half the resolution for red (1440 x 1080). But that is not accurate
either, since for most natural images the missing color pixel values can be
reconstructed very accurately, so the resolution of the red channel is
somewhere between 2880 x 2160 and 1440 x 1080.
Our goal with the D20 design is to output a very high
quality HD image with a resolution corresponding to 1920 horizontal pixels.
In order to achieve such an image output from a Bayer mask chip we need
substantially more than 1920 horizontal pixels, which is the reason the
chip's pixel count (2880 x 2160) is much higher than the desired image
output resolution. The raw Bayer data at 2880 x 2160 goes through the color
reconstruction process to fill in the missing color information and is
downscaled to a pixel count that corresponds more closely to its actual
resolution. This allows the D20 to create a high definition image that looks
as good as if not better than the images produced by current high definition
cameras.
Marc Shipman-Mueller
Courtesy of Arri
|
