Sparse Disparity Map from Uncalibrated Infrared Stereo Images

With the rapid growth in infrared sensor technology and its drastic cost reduction, the potential of application of these imaging technologies in computer vision systems has increased. One potential application for IR imaging is depth from stereo. It has been shown that the quality of uncooled sensors is not sufficient for generating dense depth maps. In this paper we investigate the production of sparse disparity maps for uncalibrated infrared stereo images, which necessitates a robust feature-based stereo matching technique capable of dealing with the problems of infrared images, such as low resolution and high noise. Initially, a set of stable and tractable features are extracted from stereo pairs using the phase congruency model. Then, a set of Log-Gabor wavelet coefficients in different orientations and frequencies are used to analyze and describe the extracted features for matching. Finally, epipolar geometrical constraints are employed to refine the matching results. Experiments on a set of IR stereo pairs validate the robustness of our technique.