Digital BW, The Print

Bob writes:

> Surely every pixel on a digital camera (except
> for a few exotica) is interpolated?

No.  Normally each pixel in the output image corresponds to a physical pixel
in the CCD.  The only interpolation that occurs is for color, since each
pixel has a color filter over it to restrict incoming light to red, blue, or
green (with green being twice as common as the other two colors, since the
eye is more sensitive to it).  So color information is significantly
interpolated, and luminance is slightly interpolated, but the actual
resolution of the output image overall matches that of the CCD.

Additionally, most interpolation does not inflate resolution artificially.
If anything, in good cameras, the resolution in the output image is _lower_
than the resolution of the CCD, because interpolation is used to produce a
smaller number of pixels of better quality.

The Fuji case, however, inflates around 6 megapixels to twice that number in
the output image, and it's just as much a case of hype as are scanners that
"interpolate" to much higher resolutions than they can actually provide
optically.  Pixel inflation in this way does nothing to improve image
quality; it only helps to sell cameras to the unwary.

> The skill of the camera maker is surely in choosing
> the best size and number of sensors, their physical
> arrangement, and the interpolation techniques.

Maybe, but interpolation to twice the resolution of the chip is never
justified for photosites of symmetrical configuration, no matter what the
skill of the camera maker.

> For example, a Nikon D1x has a chip with 4028 x
> 1324 sensors, but the image produced is 3008 x
> 1960 pixels - all interpolated !!

The difference is only 9.6%--still suspicious, but hardly the 100% different
advertised by Fuji.  The D1x has an unusual photosite configuration as well,
as I recall.

> As to 6 mpixels being all one needs, I can see
> the difference between my current scanned and
> printed images with 24 mpixels and my previous
> ones with 'only' 11 mpixels.

Not on a print held at standard viewing distance.

Under absolutely perfect viewing conditions, when viewing an image with an
aspect ratio of 3:2 at a "standard" distance equal to the diagonal (about
3.6), you can see a maximum of 21.8 megapixels.  This limit is imposed by
the dimensions and spacing of the cone cells in the retina, and it cannot
get any better than this.  However, it can and does get a lot worse: under
more normal conditions, resolving power is at least twice as bad, which
gives a total image size of 2x2 fewer pixels, or about 6 megapixels.

The only way to exceed this figure is to view the image from a closer
distance.  The only time 6-8 megapixels is insufficient, then, is when the
image will be viewed from less than the diagonal, or when it will be cropped
(the cropped portion must still have 6-8 megapixels, which means that the
uncropped version must be correspondingly larger).

Most small prints are actually viewed from distances larger than the
diagonal, and so the required number of pixels is even smaller.  For a
4x6-inch snapshot viewed from 19" away, even one megapixel is fine.

On the other hand, if you make a wall-sized 6x6-foot enlargement and view it
from two feet away, you'll need 106 megapixels (more than 10,000 pixels on a
side) in the image in order to guarantee that it looks sharp from that
distance.  This requires at least medium-format gear.

> With just 6 mpixels an A3+ print would only
> have 167 ppi!

That's all it needs, as long as you view it from at least 50 cm away.

> With 24 mpixels I can send 330 ppi to the printer
> - in the recommended range.

Inkjet printers can't print more than about 150-160 ppi; the rest is for
depth of modulation (whence the recommendation for a slightly higher
resolution).  For large A3 prints and beyond, however, at normal viewing
distances, you don't need that much resolution on the print.