Elements of information theory
Elements of information theory
SIGGRAPH '93 Proceedings of the 20th annual conference on Computer graphics and interactive techniques
SIGGRAPH '96 Proceedings of the 23rd annual conference on Computer graphics and interactive techniques
Alignment by Maximization of Mutual Information
International Journal of Computer Vision
Epipolar Geometry in Stereo, Motion, and Object Recognition: A Unified Approach
Epipolar Geometry in Stereo, Motion, and Object Recognition: A Unified Approach
Multiple View Geometry in Computer Vision
Multiple View Geometry in Computer Vision
The Amsterdam Library of Object Images
International Journal of Computer Vision
| Hi-index | 0.00 |
Piecewise-linear methods accomplish the registration by dividing the images in corresponding triangular patches, which are individually mapped through affine transformations. For this process to be successful, every pair of corresponding patches must lie on projections of a 3D plane surface; otherwise, the registration may generate undesirable artifacts, such as broken lines, which diminish the registration quality. This paper presents a new technique for improving the registration consistency by automatically refining the topology of the corresponding triangular meshes used by this method. Our approach iteratively modifies the connectivity of the meshes by swapping edges. For detecting the edges to be swapped, we analyze the local registration consistency before and after applying the action, employing for that the mutual information (MI), a metric for registration consistency significantly more robust than other well-known metrics such as normalized cross correlation (NCC) or sum of square differences (SSD). The proposed method has been successfully tested with different sets of test images, both synthetic and real.