Digital BW, The Print

Paul D. DeRocco writes:

> Well, now we're back to spatial resolution. You said color resolution.

The resolution of colors in space is spatial resolution.  Spatial
resolution in color is color resolution.

> Look, if an scene contains a red dot that happens to fall on a blue sensor
> element, you won't see it.

True.  Whereas, if you are using a sensor that captures all three colors
for each pixel, you _will_ see it.

> But you began with the statement that digital images (by which I take
> you to mean Bayer pattern images) have a certain "look" which can be
> seen, presumably without comparing the image with the original scene
> and looking for obscure flaws in the details.

And they do, which is why digital images have a "look."  Some people
like the creamy pastel look, some don't.

When the sensor is of sufficient size and quality, with appropriate
electronics and software to back it up, and when you use three sensors
or the equivalent to capture all three colors at every pixel, and when
you have a sensor with sufficiently high resolution, you can get better
results from digital than from film.  But real-world cameras at
realistic prices simply don't come anywhere near that ideal for now.
Until they do, it will be relatively easy to recognize digital photos by
their "look," at least before editing.

It's true that the absence of grain is part of the look, which should
not be a defect, but part of that absence of grain is _due_ to a defect,
in this case (namely, the absence of full color information).

> Everything to do with what?

Everything to do with the pastel look I described.  A limited gamut
produces faded-looking colors that tend to resemble each other a lot.

> The different gamuts of offset prints, chemical prints, inkjet prints,
> and so on, has nothing to do with Bayer patterns.

The limited gamut of a typical digicam has a great deal to do with
matrix filters.

> They don't remove 2/3 of the information; they remove 2/3
> of the data.

Same thing, in this case.  All light coming from the scene is
information _and_ data.

> Having a Bayer pattern does absolutely nothing to reduce the range of
> possible colors; it only means that you can't represent arbitrary
> colored one-pixel dots.

That is a serious limitation, and it does reduce the range of possible
colors overall.  Remember, 2/3 of the image is gone before it reaches
the sensor.

> But a photograph doesn't consist of abritrary colored dots, it
> consists of meaningful shapes and textures ...

Then how do we manage to represent photographs with arbitrary colored
dots?

> ... which is why we can recognize a farm in an impressionist painting.

Like the paintings of Georges Seurat?

> Modern Bayer interpolation recognizes shapes, or more precisely edges
> and gradients, and makes intelligent choices about how to fill the
> pixels in.

That still isn't the same as actually capturing the information.

> In 99% of all real-world images, it does this subjectively perfectly.

And it all cases, it does it objectively wrong.  Personally, I prefer to
see what the scene really looked like, not what the software estimates
it might have looked like.

> I don't mean to offend ...

I'm never offended.  Indeed, personal attacks work in my favor.

> ... but you make astonishing, outlandish statements that
> leave me nearly speechless sometimes.

In other words, I say things with which you don't agree, and since you
can't always think of objective arguments to support your own opinion,
you lash out with personal attacks in the hope that that will somehow
validate your own position or discredit mine.

> As you can see, I hadn't run out of arguments. I've just run
> out of patience.

None of your arguments is persuasive.  My experience is that people with
good arguments always have more than enough patience to present them.

To summarize, if you like the silky-smooth, slightly blurry pastels of
digital, that's fine; but you should not attempt to prove that a
defective design (like a matrix filter) is better than or equal to a
superior design (like three sensors of the equivalent capture of all
color data), because it is a losing battle.