No abstract

ABSTRACT:The increasing availability of high resolution stereo images from Earth observation satellites has boosted the development of tools for producing 3D elevation models. The objective of these tools is to produce digital elevation models of very large areas with minimal human intervention. The development of these tools has been shaped by the constraints of the remote sensing acquisition, for example, using ad hoc stereo matching tools to deal with the pushbroom image geometry. However, this specialization has also created a gap with respect to the fields of computer vision and image processing, where these constraints are usually factored out. In this work we propose a fully automatic and modular stereo pipeline to produce digital elevation models from satellite images. The aim of this new pipeline, called Satellite Stereo Pipeline and abbreviated as s2p, is to use (and test) off-the-shelf computer vision tools while abstracting from the complexity associated to satellite imaging. To this aim, images are cut in small tiles for which we proved that the pushbroom geometry is very accurately approximated by the pinhole model. These tiles are then processed with standard stereo image rectification and stereo matching tools. The specifics of satellite imaging such as pointing accuracy refinement, estimation of the initial elevation from SRTM data, and geodetic coordinate systems are handled transparently by s2p. We demonstrate the robustness of our approach on a large database of satellite images and by providing an online demo of s2p.Figure 1: 3D point clouds automatically generated from Pléiades stereo datasets, without any manual intervention, with the s2p stereo pipeline. Its implementation can be tested online through a web browser.

Semi-global matching [3] (SGM) is a stereovision algorithm that approximately minimizes a global energy composed of pixel-wise matching cost and pair-wise smoothness terms. The accuracy and speed of SGM are the main reasons for its widespread adoption, even for applications beyond stereovision. In SGM the two-dimensional smoothness constraint is approximated as the average of one-dimensional line optimization problems, which amounts to solving the problem on a star-shaped graph (usually with 8 cardinal directions) centered at each pixel. However, since two adjacent scan lines share little information, this approximation also produces characteristic streaks in the final disparity image (see fig.1).Based on a recently proposed interpretation of SGM as a min-sum Belief Propagation algorithm [1], we propose in this paper a new algorithm that improves the energy gap of SGM with respect to more comprehensive optimization algorithms. The proposed method comes with no compromises with respect to the baseline SGM, no parameters and virtually no computational overhead. At the same time it yields higher quality results by removing the streaking artifacts of SGM.SGM formulates stereo matching as finding the disparity map D that minimizes the global energy defined on the graph G = (I, E)where the unary terms C p (d) represent the pixel-wise cost of matchingThe pairwise terms V (·, ·) enforce smoothness of the solution by penalizing changes of neighboring disparities on the edge set E (usually the 8-connected image graph). SGM considers truncated pairwise terms of the form (with P2 > P1)In SGM the 2D problem (1) is splitted into 1D sub-problems defined on scan lines that run through the image in the 8 cardinal directions. For each direction r SGM recursively computes the costs L r from the edges of the image along the path in the direction r:The form of the smoothness potential (2) permits to compute L r (p, ·) with just 7 instructions per disparity [2]. The costs L r computed for all directions r are then added to obtain the aggregated cost volume from which the final disparity is selected with a Winner-Take-All (WTA) strategy. The SGM algorithm amounts to the min-sum Belief Propagation algorithm on a star-shaped graph centered at each pixel [1]. That is, the recursive formula (3) is actually computing the state belief of the node p for each r-oriented path. And the aggregate of state beliefs for the 8 directions (N dir = 8)corresponds to the min-marginals for the star-shaped graph centered at p. Min-sum Belief Propagation (BP) [4]can be used as an approximate energy minimization algorithm on a graph. On a generic graph, BP computes each node's belief by sending messages along the edges of the graph. A message from node q to node p is defined recursively as The state belief of a node is then computed from the messages as More Global Matching 1 (MGM). Our contribution consists in changing the recursive update formula (3). During the left-to-right pass of SGM the image is traversed in raster order (left-right, top-down), but n...

Point sets obtained by 3D scanners are often corrupted with noise, that can have several causes, such as a tangential acquisition direction, changing environmental lights or a reflective object material. It is thus crucial to design efficient tools to remove noise from the acquired data without removing important information such as sharp edges or shape details. To do so, Fleishman et al. introduced a bilateral filter for meshes adapted from the bilateral filter for gray level images. This anisotropic filter denoises a point with respect to its neighbors by considering not only the distance from the neighbors to the point but also the distance along a normal direction. This simple fact allows for a much better preservation of sharp edges. In this paper, we analyze a parallel implementation of the bilateral filter adapted for point clouds. Source CodeThe ANSI C++ source code permitting to reproduce results from the on-line demo is available on the web page of the article 1 .