Intrinsic Image Estimation and Image Deblurring for Microscopic Images

Stereo microscopes used in micron-level vision research do not directly provide the reflectivity field that we would like to measure for a number of reasons: 1) The measurement system causes blurring in reflectivities, 2) The measured image is a composition of reflectivity and illumination, 3) The measured data contain some random noise. In this work, we propose a method for finding the reflectivity image from the measured stereo microscopic data. First, we find the two intrinsic images, reflectivity and illumination, that make up the measured image. However the reflectivity component here still contains blurring and measurement noise. We use Tikhonov and edge-preserving regularization to deal with these problems, and estimate the underlying reflectivity image. We demonstrate the performance of our technique through experimental results on real data.