Tree [r78] /

File Date Author Commit
DemoSpectrum 2007-03-18 kumosan [r44] mainly docs
DocuImages 2007-03-30 kumosan [r50] a few bugfixes and code comments added
ExportFilter 2010-07-05 kumosan [r75] Second state cmake integration
FilterHandler 2010-07-05 kumosan [r78] FilterHandler cmake branch
SpectraScan 2010-07-01 kumosan [r73] Path corrections
AUTHORS 2010-05-27 annulen [r64] Added AUTHORS
CHANGES 2007-03-31 kumosan [r51] export plugin improved
CMakeLists.txt 2010-07-05 kumosan [r76] Complete cmake support
COPYING 2007-01-28 kumosan [r4] Path to version -3
README 2007-03-18 kumosan [r44] mainly docs 2007-04-18 kumosan [r54] Bugfixes and enhancements for filter plugin code
SpectraScan.dox 2007-03-18 kumosan [r44] mainly docs

Read Me

What is SpectraScan?
Currently not much more than a concept study.
A friend asked me if I knew a program, which allows to
bring scans of paper spectra (ESR,NMR,MS etc)  in jpg or another graphic 
format back into a computer readable form. Something like an ocr program for 
spectra. I did not know of such a program.

Initially I thought this would be impossible or at least
a tremendous task. Too hard for some hobby projects after work or during 
weekends. But after some thought, it actually seems to be terribly easy. 

Load an Image, cut off everything that is not curve, and
write out the x,y-coordinates of all black pixels in CSV format.
Thats all. Load the coordinates in a statistic program of your
choice and do a curve fit. If the scan isn't too bad (usually they are), 
the fit might not be necessary and the curve could be loaded directly into 
a spreadsheet program. 

This is what SpectraScan does. It just creates CSV files from graphic files..

I gave this initial version even a minus number. Ignore the
GUI, ignore the code. At this time it was only a 'wrapper' to
get a few CSV files. Just to see if the idea really works.
It does. Now the real development can start. 


Scanned images often have 'dirt points' all over a spectrum.
Development of a few filters, which delete non-curve points,
which mess with the curve fit.

Currently all scale information gets lost. This is not really
a problem, since this could be easily fixed in most programs which are 
suitable to do the final curve fits. However, to keep the scale information 
in the CSV coordinates is really easy and makes the handling much smoother.




Don't know. Depends on what problems might appear. Currently I tested only 
one spectrum. And of course it depends on the feedback and wishes I get.

Its own curve fitting routines, making if a fully capable application? 
Who knows.


SpectraScan is a Qt 4.2 program. It definitely works
under Linux and it has been successfully compiled with Visual Studio 2005.

For Linux the usual Qt stuff:



This creates the main program + all currently available plugins.
Just move the binary whereever you want it. Copy the Plugin folder in the
same place where you put the binary.

Of course, this is not the final installation routine.
But for the time being, i.e. until SpectraScan reaches 
beta status, it must suffice.