Mosaicing with Parallax using Time Warping

2D image alignment methods are applied successfully for mosaicing aerial images, but fail when the camera moves in a 3D scene. Such methods can not handle 3D parallax, resulting in distorted mosaicing. General egomotion methods are slow, and do not have the robustness of 2D alignment methods. We propose to use the x-y-t space-time volume as a tool for depth invariant mosaicing. When the camera moves on a straight line in the x direction, a y-t slice of the space-time volume is a panoramic mosaic, while a x-t slice is an EPI plane. Time warping, which is a resampling of the t axix, is used to form straight feature lines in the EPI planes. This process will simultaneously give best panoramic mosaic in the y-t slices. This approach is as robust as 2D alignment methods, while giving depth invariant motion ("ego motion"). Extensions for two dimensional camera motion on a plane are also described, with applications for 2D mosaicing, and for image based rendering such as "light field".

[1]  Xueyin Lin,et al.  Panoramic EPI generation and analysis of video from a moving platform with vibration , 1999, Proceedings. 1999 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (Cat. No PR00149).

[2]  Richard I. Hartley,et al.  Theory and Practice of Projective Rectification , 1999, International Journal of Computer Vision.

[3]  Katsushi Ikeuchi,et al.  ENHANCED NAVIGATION SYSTEM WITH REAL IMAGES AND REAL-TIME INFORMATION , 2001 .

[4]  Richard Szeliski,et al.  The lumigraph , 1996, SIGGRAPH.

[5]  Allen R. Hanson,et al.  Parallel-perspective stereo mosaics , 2001, Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001.

[6]  Marc Levoy,et al.  Light field rendering , 1996, SIGGRAPH.

[7]  Richard Szeliski,et al.  Construction of Panoramic Image Mosaics with Global and Local Alignment , 2001 .

[8]  Daphna Weinshall,et al.  Mosaicing New Views: The Crossed-Slits Projection , 2003, IEEE Trans. Pattern Anal. Mach. Intell..

[9]  Harry Shum,et al.  Construction and refinement of panoramic mosaics with global and local alignment , 1998, Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271).

[10]  K. Hanna Direct multi-resolution estimation of ego-motion and structure from motion , 1991, Proceedings of the IEEE Workshop on Visual Motion.

[11]  Harpreet S. Sawhney,et al.  Robust Video Mosaicing through Topology Inference and Local to Global Alignment , 1998, ECCV.

[12]  Robert C. Bolles,et al.  Epipolar-plane image analysis: An approach to determining structure from motion , 1987, International Journal of Computer Vision.

[13]  Richard Szeliski,et al.  Handling occlusions in dense multi-view stereo , 2001, Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001.

[14]  P. Anandan,et al.  Hierarchical Model-Based Motion Estimation , 1992, ECCV.

[15]  Takeo Kanade,et al.  A Multiple-Baseline Stereo , 1993, IEEE Trans. Pattern Anal. Mach. Intell..

[16]  Hiroshi Kawasaki,et al.  Ego-Motion Estimation for Efficient City Modeling by Using Epipolar Plane Range Image Analysis , 2003 .

[17]  Allen R. Hanson,et al.  Error Characteristics of Parallel-Perspective Stereo Mosaics , 2001 .

[18]  O. Faugeras Three-dimensional computer vision: a geometric viewpoint , 1993 .

[19]  P. Anandan,et al.  Direct Recovery of Planar-Parallax from Multiple Frames , 1999, IEEE Trans. Pattern Anal. Mach. Intell..

[20]  Shmuel Peleg,et al.  Mosaicing on Adaptive Manifolds , 2000, IEEE Trans. Pattern Anal. Mach. Intell..