Abstract
We present a Generalized Deformable Spatial Pyramid(GDSP) matching algorithm for calculating the dense cor-respondence between a pair of images with large appear-ance variations. The main challenges of the problem gen-erally originate in appearance dissimilarities and geometric variations between images. To address these challenges,we improve the existing Deformable Spatial Pyramid (DSP)[10] model by generalizing the search space and devisingthe spatial smoothness. The former is leveraged by rota-tions and scales, and the latter simultaneously considersdependencies between high-dimensional labels through the pyramid structure. Our spatial regularization in the high-dimensional space enables our model to effectively pre-serve the meaningful geometry of objects in the input im-ages while allowing for a wide range of geometry variationssuch as perspective transform and non-rigid deformation. The experimental results on public datasets and challeng-ing scenarios show that our method outperforms the state-of-the-art methods both qualitatively and quantitatively. F