A SCALE SPACE FOR TEXTURE+DEPTH IMAGES BASED ON A DISCRETE LAPLACIAN OPERATOR Anonymous ICME submission

In this paper we design a smoothing filter for texture+depth images based on anisotropic diffusion. Our proposed filter is linear and allows to generate a scale space of the texture image which preserves internal image structure (i.e. it does not smooth through object boundaries). Moreover, we show experimentally that our filter demonstrates better stability under camera position changes in such a way that its response is rather intrinsic to the content and less dependent on the observer position in the scene. To illustrate the practical utility of the scale space with such properties, we integrate our filter into SIFT salient visual point detector, getting a substantial improvement of the repeatability of detected keypoints under significant viewpoint position changes.