Depth maps upsampling techniques recently debated in technical literature involve statistical, possibly Markovian, priors [24, 25], as well as video sequence analysis [37, 38]. In Wang et al. [37, 38], the depth map interpolation method exploits the existing pairs of key frames and their corresponding depth maps. The method requires frame selection, motion alignment, and updating and adjusting of the map. In Lee and Ho [21] the authors generate multi-view video sequences and their corresponding depth maps by making use of a multi-camera system enriched by one time-of-flight depth camera. An initial virtual depth map is generated at each viewpoint, and then the depth map edges are adapted to the estimated moving object edges. The work in [22, 23] deals with adaptive edge-preserving depth map smoothing, which counteracts typical depth based image rendering artifacts, ie holes on the generated virtual view images. In Zhu [42],method for combining the results from both active and passive techniques is built on the base of a pixel-wise reliability function computed for each acquisition method. In Jaesik et al. [19] high quality upsampling uses an additional edge-weighted nonlocal means filtering to regularize depth maps in order to maintain fine detail and structure. The depth image upsampling method in [16] exploits a novel edge layer concept and encompasses it in the design of a non-linear interpolation filter for depth map up-sampling. Edge preserving upsampling for depth map coding purposes appears in [26]. In Tseng and Lai [35], the depth map for all pixels at each frame is estimated by first applying correspondence search, computing the optical flow between selected scenes, and then refining the results by MRF. A depth map post processing scheme for FVT applications appears in [3]. In Xu et al. [40] the depth map preprocessing method using texture edge information to refine depth pixels around the large depth discontinuities. The method in [31] uses edge information from the video frame in the upscaling process and allows the use of sparse, non-uniformly distributed depth values. The method in Rong et al. [28] addresses sharpening edges of the depth maps possibliy degraded by coding. In Vijayanagar et al. [36] a post-processing technique corrects and sharpens the boundaries of the objects of the depth map and ensures local depth smoothness within the objects. The work in De Silva et al. [11] exploits the depth perception sensitivity of humans in suppressing the unnecessary spatial depth details, hence reducing the transmission overhead allocated to depth maps. In Ekmekcioglu [12] an edge adaptive depth map upsampling procedure stemming from an edge map of the original natural image has been proposed. The resulting upsampled depth maps exhibit an increased sharpness in edge regions with respect to classical upsampling such as bilinear interpolation and linear spatial scalable filters [33]. Since no information about the orientation of the edges is considered, some edges still exhibits visual artifacts which, in turn, may severely affect all of the stages where a correct high resolution depth map is required, such as, for instance, the rendering of free viewpoint images. Moreover, the work in Ekmekcioglu [12] estimates the edge map from the high resolution luminance map, thus needing the samples of the low resolution depth map to be correctly registered with the sample of the high resolution luminance map. This latter operation is computationally onerous; besides, it may affect the upsampling procedure when not perfectly performed (Table 1).