An imaging variable retardance polarimeter was developed and tested by Tyo
and Turner [Proc. SPIE 3753, 214 (1999)]. The signal-to-noise ratio (SNR) i
n the reconstructed polarization images obtained with this system varied fo
r the four Stokes parameters. The difference in SNR is determined to be due
to differences in the Euclidean lengths of the rows of the synthesis matri
x used to reconstruct the Stokes parameters from the measured intensity dat
a. I equalize (and minimize) the lengths of the rows of this matrix by mini
mizing the condition number of the synthesis matrix, thereby maximizing the
relative importance of each of the polarimeter measurements. The performan
ce of the optimized system is demonstrated with simulated data, and the SNR
is shown to increase from a worst case of -3.1 dB for the original setting
s to a worst case of +5.0 dB for the optimized system. (C) 2000 Optical Soc
iety of America OCIS codes: 260.5430, 230.5440.