Digital imaging radiometry and photometry
using raw data formats of common cameras

is possible thanks to the programme dcraw which reads such data and stores 
them in standard formats.

Full information can be stored if the output has not less bits then the 
raw data themselves. The most common jpeg images use just 8 bits, as well as 
the monitors etc, but most cameras have 10 or 12-bit A/D converters, and 
the original more-than-8 bit image can be used to get reliable 
information on the luminances of the imaged scene. 

All common image formats have either 8 or 16 bits per colour, and so 
the full data from cameras are to be stored rather redundantly as
16-bit ones.

Most cameras record the image through a colour grid, so that full 
colour information is contained not in a single pixel, but in a group 
of at least four of them. The standard way of constructing the image is 
computing the remaining colours for each pixel by interpolation from 
the surrounding ones. 

For measurement purposes, no interpolation is desirable, the original 
data are better. 

dcraw can store almost such data when a -d (document) option is used. 
A minor drawback is, they are stored as gray RGB data, whereas a 
single colour (gray) should do, reducing the size of the file thrice.
No interpolation is made, just autoscaling of different colours 
is made, what is suitable for the case of monochrome scenes to equalize 
the signal from neighbouring pixles of different colours.

To get scientific data enabling RGB radiometric intepretation, it is 
desirable, in a standard situation, to turn off the autoscaling, and
using either no scaling at all (to avoid overflow for at least some 
colours) or a standard predefined scaling like that giving
R=G=B for sunlit white surfaces. Autoscaling applied on such surfaces,
on the other side, may help to refine the coefficients for such
standard scaling.

One scaling is low-level one in dcraw, namely multiplying the data by a 
power of 2 to convert them to apparent 14 bit ones. In this way, 
overflow should begin near to 12000, for no more scaled data.

Apart from scaling, the unlit pixel mean value is subtracted (like dark 
current for classic photoeletric photometry or as a substitution of 
subtracting a really dark image from the real image). This destroys 
some information in pixels which are below this value (all such become 
zero), representing either information on the noise or even real data 
(for low-noise pixels). Perhaps subtracting some value less by ten
or fifty would prevent most of that that loss. 

The best way would to subtract no constant, but really a 
dark frame, again less by some small value (like 4) to prevent leveling
at zero. Like in the standard scientific imaging photometry.  Division
of the resulting frame by a similarly processed  white frame
(``flatfield'') to equalize the sensitivities of the  pixels would
enable even a reliable point-source photometry  (otherwise, any single
pixel is unreliable, just at least dozens  of them can be taken as a
measure of the amount of light). (Still, because of the colour matrix,
to measure points, they should be defocused to  cover at least four
neighbouring pixels.)

The portable graymap *.pgm (as well as portable colourmaps, *.ppm) can 
have comments on separate lines (beginning with #) after the initial data 
(P5 width height ) and before the maxval information. 
In the comments, all informatin on black level mean or
median value (or even on some  percentils), really subtracted value,
scaling of colours, all-pixel scaling etc. could be mentioned. 
Using such information, the  exact RGB
radiances and luminance can be computed for each frame, when
calibration images exists (white paper lit by a sun high in
the clear sky etc.).  

Such an image could be then converted to four FITS files for processing 
with some standard astronomical software, esp. for sake of photometry 
of point-like sources.



I've adapted the Dave Coffin's Raw Photo Decoder v4.86 in such a 
manner, that a new option -9 produces a pgm file with all such 
comments. 

Even 8-bit interpolated ppm files have the needed comments in this
version, and if they have no gamma<>1, their use for photometry is
possible as well (with the advantage of a possibility  to use the most
common screen photo editors and still arrive at  luminances or RGB
radiances). 

jenik hollan