Post-calibration row- and column-correlated noise resulting from voltage instability in the electronics of the Thermal Emission Imaging System (THEMIS), on-board the 2001 Mars Odyssey spacecraft is mapped with the data during map projection, making reliable spectral analysis of small areas (< 4km2) problematic. The majority of detector-row and column correlated noise is removed during calibration using a windowed algorithm and utilizing information in the Instrument Response Function (IRF), but residual noise dubbed 'plaid' and processing artifacts remain after the calibrated data has been projected. These artifacts can be readily viewed using decorrelation stretches or by numerically comparing radiance or emissivity differences in small regions. We present an algorithm to remove the residual row- and column-correlated noise from THEMIS data, leveraging post-projection information to improve the accuracy in radiance and emissivity analyses of small areas. The removal of the 'plaid' is essential for the detection of spectrally significant surface materials.