Scarse (Scanner CAlibration ReaSonably Easy)  --  readme
========================================================

What is it?
===========

Scarse is a free color calibration and management software package. It
lets you build and use ICC profiles. Custom profiles can be generated
from variety of calibration targets. Scarse is intended for (and
developed on) Unix machines and is distributed under the terms of GNU
Public License (see file COPYING).

Scarse project was born out of my desire to be able to get good scans
which are consistent across large batches, and do it with minimal
amount of manual tweaking. It is intended for serious photographic or
prepress applications, and is most useful with high-end film scanners.
If you just want to occasionally scan a snapshot of your aunt on your
$100 flatbed and don't care about the colors as long as they are
snappy, don't bother, this package is not for you.

As the name implies, color management issues are rarely addressed in
the free software world, probably because until now good scanner
hardware was unaffordable to a mere mortal. With all professional
graphic solutions being firmly locked into Mac or Windows platforms,
there was hardly anything at all for Unix. However, time marches on,
desktop scanners and printers got much better and cheaper, some brave
people wrote Gimp, and a color management system for Unix was needed.
So I set out to write it...

Right now, the Scarse is in early development stages, and is not as
polished as commercial systems are. Indeed, it is quite rough and even
downright inconvenient in places. However, it is quite functional
already, and produces transparency scans which are superior to what I
get from OEM Polaroid SprintScan 4000 profiles (using the same
hardware). So if you are adventurous type, give it a try, and tell me
what you think...


What is it not?
===============

Scarse isn't any of the following:

  * Scanner or printer driver:
      Scarse knows nothing about your scanner and printer hardware. It
      relies on external drivers to do actual scanning and printing. If
      you are looking for scanner drivers, you want SANE.
  * Image manipulation program:
      Scarse does not do anything to the image itself. It just
      manipulates colors so they appear correct on your hardware. If
      you want to retouch photos or apply unsharp masks and such, you
      need Gimp.
  * ICC engine:
      Scarse isn't an ICC engine per se. It builds on top of icclib,
      which is an ICC engine. Scarse is a color management system,
      which means it allows you to generate custom profiles for your
      hardware setup, as well as use them.
  * Fast:
      Maximum possible speed was never a design goal. The precision of
      color translation is the priority. Scarse uses doubles for color
      representation internally, which gives you full 52 bits/channel,
      as opposed to only 8 or 16 bits/channel in packed integer
      representation you typically see. This eliminates discretization
      errors, but makes it run slower (still, it is not a slug :).


How does it work?
=================

It works quite well, thank you :). Oh, you meant technical details :)? 

Scarse works by comparing the scan of a hardcopy reference calibration
target (such as IT8.7 or Kodak Q60) coming from your hardware to the
actual colors the target is supposed to have. It then builds (by
approximation and interpolation) lookup tables that take into account
the measured scanner hardware peculiarities. Using thus generated
profile, you can then get scans with color rendition (more or less :)
independent of the particular hardware used. 

You will need a supported hardcopy calibration target. Standard IT8.7
targets are made by most major film manufacturers for the purpose of
calibrating scanners for their films. They offer a lot of color points
to measure and thus produce quite precise profiles. Q60 target is
available from Kodak in 35mm and 4x5 transparency, as well as 5x7
reflective formats. Wolf Faust makes 5x7 reflective IT8.7 targets on
Fuji paper, which he sells quite cheaply. Other standard calibration
targets, such as MacBeth Color Checker and Kodak Color Separation Guide
Q13 & Q14 (grayscale only), are also supported.

You cannot calibrate scanner for negative film the same way you can do
it with slides, because of the variations in negatives on per-exposure
basis. Scarse provides automatic color adjustment which will normalize
scan of a negative to roughly correct levels and contrast, and you can
work from there. Printing negatives is more of a subjective thing, and
some decisions are better left to a human.


Where do I find it?
===================

The latest release and development news can be found at Scarse homepage

	http://www.scarse.org/

You can contact the author (that's me :) directly by email

	Andrei Frolov <andrei@phys.ualberta.ca>


What's included?
================

Included in this distribution are:
	
	CHANGES		- change log
	COPYING		- GNU Public License
	CREDITS		- people who contributed to development
	INSTALL		- installation notes
	README		- this file
	TODO		- to do list
	VERSION		- distribution version number
	
	icclib/		- icclib & examples by Graeme W. Gill
	data/		- calibration target data and so on
	src/		- scarse source code, including
	  ipb		    ICC profile builder
	  calibrate	    input/output device calibration utility
	  cmap		    utility to translate image through ICC profiles


How do I compile/install/use it?
================================

Read installation notes in file INSTALL - in brief, edit makefiles to
suit your machine configuration and then say

	make; make install

in the top directory.

As for using it, I have not written proper documentation and man pages
yet, so you are on your own :). Tutorial on color calibration could not
hurt either, as there is a lot of confusion around the issue. But for
now, I'll assume that you know what you are doing, and are not afraid
of reading source code :). So I'll just give a few examples of what you
can do with this package. Note that all tools will print the summary of
their usage and options if invoked with '-h' flag.

To generate scanner profile:

    Scan calibration target (Kodak Q60) at full bit width hardware
    support, and save it as TIFF file. Open it up in viewer and find
    the position of the target corners (those white/black crosslines
    around the target grid) and write it down as X geometry
    specification in percent of the total image size. You can also
    just crop the image to the target corners, and skip the geometry
    specification. (If you are unclear on what I mean, see example at
    Scarse home page.) This step is important, as calibration utility
    must align grid with all those colored squares, or else it will
    produce bogus colors. Then say something like

        calibrate -v -d scanner -tQ60E3 -b1998:08 -g 90x80+5+10 -i file.tif profile.icm
    
    If it complains about 'fluctuations too big', either your scan is
    really noisy, or the grid was misaligned. Check the geometry
    specification again if unsure.

To generate display profile:

    Proper display calibration is not supported yet (and would require
    some kind of photometer, anyway). However, you can easily generate
    generic power-law profiles by saying something like
    
        ipb -v -c display -iRGB:2.5 -p Trinitron -p D50 -M profile.icm
    
    The trick is to get gamma and RGB primaries right. Typical gamma
    for PC monitor is 2.5, for Mac it is 1.8. Scarse supports all the
    primaries Photoshop knows about (and you can specify your own too),
    but you will have to choose the ones you want yourself :). White
    point should probably be 5000K (D50) regardless of RGB primaries
    you choose (to match with scanner profiles), but you can override
    that too.

Display vs. working space:

    RGB spaces based on monitor primaries have somewhat limited color
    gamut. If you intend to go to press, or want to make an archival
    copy of your photos, you may want to work in a different color
    space. Lab has the largest gamut, but is currently not supprted by
    Gimp at all. Wide Gamut RGB or Adobe RGB could be viable choices.
    In any case, corresponding profiles can be generated by choosing
    appropriate primaries with -p option to ipb:

        ipb -v -c display -iRGB:2.2 -p Adobe -p D50 -M profile.icm

To generate generic profile:

    ICC profile builder, ipb, can build profiles connecting various
    color spaces, generic or tailored to fit some set of measured data.
    For example, to generate generic XYZ to Lab transform, say

   	ipb -v -c abstract -iXYZ -oLab profile.icm

To use the profiles you generated:

    Use cmap to translate image through a series of profiles, and store
    the result as TIFF image. You can also control how image is stored,
    check cmap options for more details. Typical use for scans of
    positive media (transparency or reflective) would be

        cmap -v -p scanner.icm -r display.icm raw.tif calibrated.tif

Handling negative scans:

    You cannot calibrate scanner for negative film the same way you can
    do it with slides, because of the huge variations in negatives on
    per-exposure basis. To work around this, automatic color adjustment
    algorithm has been encorporated into cmap. You use it like this
    
        cmap -v -ALg raw.tif corrected.tif
    
    The automatic correction brings the image in the rough ballpark,
    and you can further manually correct levels and curves - see cmap
    help. Some films (notably 4-layer emulsions like Fuji NPS) seem to
    like more aggessive per-channel automatic correction given by -ALG.


What if I find a bug?
=====================

What do you mean, a bug? It's a feature! :) Seriously though, as this
is first alpha, and not tested to any significant extent, expect bugs.
Lots of them. If you find one, please send bug report to me (my email
is at the bottom of this page), including description what went wrong
and what were you doing at the time. Better yet, fix it yourself and
send me a patch :). I also want to hear improvement suggestions and
(constructive) critique.

And if you want to contribute to this project, please do!



Well, that's all for now...

-- Andrei Frolov <andrei@phys.ualberta.ca>