Introduction to algorithms
Matching Two Perspective Views
IEEE Transactions on Pattern Analysis and Machine Intelligence
Robust Egomotion Estimation From the Normal Flow Using Search Subspaces
IEEE Transactions on Pattern Analysis and Machine Intelligence
Computer and Robot Vision
Stereo Matching in the Presence of Narrow Occluding Objects Using Dynamic Disparity Search
IEEE Transactions on Pattern Analysis and Machine Intelligence
ECCV '96 Proceedings of the 4th European Conference on Computer Vision-Volume I - Volume I
| Hi-index | 0.00 |
Stereo correspondence is a central issue in computer vision. The traditional approach involves extracting image features, establishing correspondences based on photometric and geometric criteria and finally, determine a dense disparity field by interpolation. In this context, occlusions are considered as undesirable artifacts and often ignored. The challenging problems addressed in this paper are a) finding an image representation that facilitates (or even trivializes) the matching procedure and, b) detecting and including occlusion points in such representation. We propose a new image representation called Intrinsic Images that can be used to solve correspondence problems within a natural and intuitive framework. Intrinsic images combine photometric and geometric descriptors of a stereo image pair. We extend this framework to deal with occlusions and brightness changes between two views. We show that this new representation greatly simplifies the computation of dense disparity maps and the synthesis of novel views of a given scene, obtained directly from this image representation. Results are shown to illustrate the performance of the proposed methodology, under perspective effects and in the presence of occlusions.