Re: Why So Many "RAW" Formats?


On Tue, 10 Feb 2009 18:31:34 -0900, floyd@xxxxxxxxxx (Floyd L.
Davidson) wrote:

TheRealSteve <steve@xxxxxxxxxxx> wrote:
On Tue, 10 Feb 2009 05:05:53 -0900, floyd@xxxxxxxxxx (Floyd L.
Davidson) wrote:

TheRealSteve <steve@xxxxxxxxxxx> wrote:
On Mon, 09 Feb 2009 20:34:10 -0900, floyd@xxxxxxxxxx (Floyd L.
Davidson) wrote:

C J Campbell <christophercampbellremovethis@xxxxxxxxxxx> wrote:
On 2009-02-09 17:30:14 -0800, floyd@xxxxxxxxxx (Floyd L. Davidson) said:

ray <ray@xxxxxxxxxx> wrote:
On Mon, 09 Feb 2009 07:30:15 -0800, Matt Ion wrote:

Bob Williams wrote:
Seems like every Camera manufacturer has one or more of his own
versions of a RAW image format. And many (most?) versions are
proprietary.
RAW, by its very definition, is not an "image
format" but simply the raw
data straight off the camera sensor. Thus any change in the sensor -
including going to a higher resolution - results in a different RAW
output.
raw data files, as any other files, have a format.
But it is not an "image format".

Sure it is. It stores an image. There is no intrinsic difference
between a RAW file and any other image format -- it is all 1s and 0s
storing an image. In fact, Nikon calls the NEF files an "image format"
and compares them to JPG and TIFF.

Nikon Electronic Format is *not* a raw image format.
Nikon often refers to "NEF raw image data", but near as
I can tell they never call NEF a "raw image format".
That is true because, if for no other reason, NEF is
used for many very different types of raw data.

And there *is* an intrinsic difference between a data
file that contains raw sensor data and an image file
which contains the data for one specific image. The raw
data file *must* be interpolated to produce an image,
and there is no one single way to do that, which means a
raw data file's data can be interpolated many different
ways. It is not data for one image, but rather data
from which an infinite number of very different images
can be made.

By your definition, there is no such thing as an image file. Even
jpegs will look very different on each monitor they're displayed on or

That has nothing to do with it. A JPEG format image
defines just one image. It might well display
differently, and it can also be edited... but it is
still just one image. There is no single image defined
by the raw data from a DSLR sensor. It is a set of data
that defines many images, not just one. And there are
many ways to get an image from the data too.

It defines one image just as much as a jpeg file defines one image. Of
course, neither one defines just one image. For instance, there is no
definition in a jpeg file for what values at each pixel equate to what
luminosity. So a single jpeg file defines an infinite number of
images

That is not true.

on each printer they're printed on. That's because processing,
including interpolation among other processing, *must* be done to a
jpeg in order to produce an image you can see on any specific medium

There is no interpolation necessary with a JPEG image.
And the processing to convert a JPEG image to some other
format is well defined to produce a *single specific*
image. Granted that it might no always actually produce
exactly the same image, but the difference is an *error*!

No it's not an error when you get differences, because there is no
single defined way to do a conversion of a jpeg file to an image.

There are multiple ways, but each is *supposed* to
regenerate the original image. The difference between
the original and what is regenerated is "error". For
example, that is one reason that JPEG is called a
"lossy" format.

I'm not even talking about the "lossiness" of the jpeg format. So
take a TIFF or a GIF or a BMP, none of which are lossy formats and all
of which define an infinite number of images from the same image file.
That's because the file does not define well enough what the final
image will look like. It will look different on each monitor and each
printer, none of which may be an "error" because the final result is
just not specified in the file. It will look different even on the
same monitor with different settings and again, it's not an error and
it's not lossy. It's just not defined in the file how bright the
monitor should be set to properly view it.

and there is no single way to do that, which means a jpeg file's data
can be interpolated many different ways.

False. JPEG data is not interpolated.

It is if you don't print or display it at it's native resolution. But
it's not just interpolated. It's also converted to light at a certain
amplitude and wavelengh at each pixel. And there's no single way
defined to do that.

If you resize it, the data is interpolated. But of
course that is also making an entirely *different* image
that the JPEG data described. (Differences in pixel
values are the error discussed above.)

That's why I'm not just talking about resizing it. I'm talking about
the infinite ways of display the *same* jpeg, gif, or tiff file...
each looking different but none of wich are an *error* because the
file doesn't specify enough about how the image should actually look
to a viewer. How much ink of what type goes where? How bright should
a pixel be lit for a given RGB value? It's just not defined. So
there are infinite ways of displaying a given jpeg, gif, bmp, tiff
etc. image file. Just like raw.

Each pixel in a JPEG is encoded into a single set of RGB
color values. That data is not used for multiple pixels
except within blocks of identical pixels.

And how do those RGB color values get display as a image? Undefined.
If you turn up or down the brightness of your monitor, you're looking
at a different image from the same jpeg.

See "error".

It's not an "error" if it's not definied. If you like viewing your
monitor set to 75% brightness and your coworker likes viewing it at
74% brightness, which one of you is in "error"? Which setting is the
*correct* setting for viewing a particular jpeg file? How does the
file specify that? Answer: it doesn't.

Raw sensor data is not a one to one relationship with
image pixels. The value eventually calculated for any
single pixel location is *interpolated* from multiple
sensor locations (and multiple pixels, each of which
might be different, use data from any single sensor
location).

Raw sensor data is a recording of an image.

No, it is recording sensor data.

Exactly, which is a way of specifying image data. The sensor data is
the data that the sensor output when it captured a certain image. If
you had a monitor with the same pattern of colored dots as the sensor
that captured the image, you wouldn't have to interpolate the sensor
data into RGB pixels. When you convert a raw sensor file to a jpeg,
all you're doing is converting one image file to another image file
because your monitor/printer can't handle the image format contained
in the raw image file directly. That doesn't mean it's not possible
to design and build one that can. It just wouldn't be very cost
effective.

There's many different
ways to interpolate it to form a different image.

At which point yes you do have an image. And those are
save to image formats such as TIFF, PPM, JPEG, etc.

You have an image before also. Just like when you interpolate a jpeg
image to form another image of a different size. Just because you can
interpolate one image to form another image doesn't mean the original
wasn't an image.

But it's still an
image. It just wouldn't look like the final result you see printed or
display. Just like a film negative is an image that doesn't look like
the final result printed.

But a film negative is just one image. The same as JPEG
is just one image.

But raw sensor data is not just one image. It is data
from which many different images could be generated.

A film negative is just like sensor data in that both are an image
from which many different images could be generated. I take it you've
never worked in a darkroom printing your own negatives and played
around with different enlargements, exposures, etc.

In fact, a raw file is even more an image file format than a jpeg file
because it is an actual photograph, i.e., a recording of photons that
struck a specific area over a specific time. Once you start messing
with that via whatever processing you choose, you're getting further
and further away from a photograph although it's still an image.

That is silliness.

Silly, but true.

If you believe that then there is no point in discussing
technical details of digital imaging with you.

If you don't know what a photograph is, i.e., a graphical
representation of the photons hitting a certain area over a certain
time, then there's no point in discussing technical details of digital
imaging with you. And for the purposes of this discussion, a file
that contains the details of a photograph (i.e., a raw sensor file)
also contains an image. Of course that photograph/image contained in
the raw sensor file can be manipulated to form an infinite number of
other images. Just like any other image file can.

Also by your definition, a film negative is not an image because it
must be printed to create an image and there's an infinite number of
ways any single negative can be printed. But I contend that a film
negative *is* an image format, just like a raw file.

Correct, a film negative is an image.

By admitting that a film negative is an image, you're being internally
inconsistent with your ascerting that a raw file is not an image file.

Not even close.

Closer than you realize.

Steve
.

Relevant Pages

  • Re: RAW or not RAW? :)
    ... When you "view" a "raw ... the program either looks at the embedded JPEG ... not converted to a standard image file format at all. ... the specific conversion process for jpeg is different than nef, ...
    (rec.photo.digital.slr-systems)
  • Re: RAW images...
    ... Saved in any non-lossy format (RAW, TIFF, ... PPM, etc.), and that is not a problem. ... using solid colors) and save it to JPEG format. ...
    (rec.photo.digital)
  • Re: RAW images...
    ... Saved in any non-lossy format (RAW, TIFF, ... using solid colors) and save it to JPEG format. ...
    (rec.photo.digital)
  • Re: RAW or not RAW? :)
    ... When you "view" a "raw ... the program either looks at the embedded JPEG ... *is* converted to an image format for viewing. ... pixel map is a "standard file format" too. ...
    (rec.photo.digital.slr-systems)
  • Re: Why So Many "RAW" Formats?
    ... versions of a RAW image format. ... It defines one image just as much as a jpeg file defines one image. ... There is no interpolation necessary with a JPEG image. ...
    (rec.photo.digital)