ones appear 
o a), d) and 
ification of 
yetween the 
k specimen 
layer a loss 
| a standard 
rge number 
ited fraction 
o for a stack 
ally for fine 
numb it is 
) and e) are 
b), but are 
20pm and 
ve, (20% for 
ss than 5% 
umes under 
caning at all 
almost none, 
: magnitude, 
anged. This 
ion behavior 
legrade the 
unted for, if 
A general 
udo-coded as 
n mean & 
ngth) 
itation) 
)6 
  
(scatter, attenuation, bleach, excitation); 
UpDateCounters 
(scatter, attenuation, bleach, excitation); 
AmplifyImageDataBy 
(scatter, attenuation, bleach, excitation); 
Adjust 
(PSF) 
END 
This reconstruction algorithm fails due to the 
unavailability of the required data and due to 
the enormous computational effort which is 
needed for each UpDateCounters operation. 
Each update step comsumes computations 
proportional to the number of voxels in the 
image data set. 
Consequently QUASIA3D solves the 
reconstruction problem with two different 
approaches. The first one implements efficient 
models and simplifications for the various 
effects in order to complete the restoration 
process in an acceptable amount of time. 
The second task is to measure the missing 
parameters needed for the restoration. 
QUASIA3D implements a 3 stage model to 
gather these data sets: 
1) The parameters of the confocal microscope 
such as wavelength sensitivity, focal spot 
geometry, PSF and objective features like 
distortion and local brightness response are 
collected via the sampling of phantom 
specimens. 
2) Dye properties are measured with the help of 
pure solutions. Excitation attenuation and 
bleaching sensitivity are estimated. To gauge 
the bleaching sensitivity as a function of 
temperature and pH (environmental 
influence), the specimens have to be 
prepared accordingly. In a standard setup, 
the temperature is kept constant and 
therefore no temperature dependency has to 
be known. 
317 
3) Specimen properties are very difficult to 
separate in a microscope. If no external 
devices can be applied to distinguish 
between scattering — and attenuation, 
experiments with fluorescent microspheres 
of precisely known shape have to be 
performed. The scattering effect shows 
frequency proportional strength. Thus 
scanning with different wavelength allows to 
differentiate between spherical aberration 
and the scattering effects. Attenuation 
versus scattering and aberration can be 
discriminated through the use of different 
aperture sizes. With high apertures 
attenuation becomes a more dominant effect. 
For a standard setup, no discrimination at all 
is required. The mixture of aberration, 
scattering and attenuation may be unified in 
an absorption term, which is typical for one 
excitation wavelength and one tissue type. 
QUASIA3D uses this approach to build up 
its specimen database. Various image 
segmentation algorithms support the user in 
identifying the different tissues and gather 
absorption data for these in a semi-automatic 
way. 
Obviously QUASIA3D is neither interactive 
nor thoroughly automated. The calibration 
procedures are semi-interactive in the sense that 
the operator must identify a set of representative 
specimen samples for the 3rd stage, but the 
reconstruction process runs without interaction 
solely based on the acquired data. This is very 
desirable, since the execution time for a 
reconstruction of an extended image stack takes 
up to 15 minutes - despite of all speed-up 
algorithms. 
6. RESULTS 
Worst case mathematical simulations which 
take all discussed effects (including cover 
glasses related) into account predict a 34% error 
bound for the acctual dye concentration 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996