Excitation light leakage through fluorescence filters limits the sensitivity of fluorescence tomography. We demonstrate an image processing approach for suppressing the effects of excitation light leakage from NIR fluorescence images. Multiple excitation and fluorescence emission wavelength images of a fluorescence target buried in 1 % Liposyn solution were acquired by scanning an intensity modulated and line shaped NIR excitation source over the illumination/detection surface of the tissue phantom. Excitation images were used to identify the NIR source pattern and multiple global and local filters were designed to mitigate excitation light leakage in the fluorescence images. The image correction approaches were assessed by comparing the processed fluorescence images with the output of an adaptive finite element based photon diffusion model. The results show that localized error modeling based on the median reduces leakage effects by about 44 % compared to uncorrected data.