We present a method to determine 3D motion and structure of multiple objects from two perspective views, using adaptive Hough transform. In our method, segmentation is determined based on a 3D rigidity constraint. Instead of searching candidate solutions over the entire five-dimensional translation and rotation parameter space, we only examine the two-dimensional translation space. We divide the input image into overlapping patches, and, for each sample of the translation space, we compute the rotation parameters of patches using least-squares fit. Every patch votes for a sample in the five-dimensional parameter space. For a patch containing multiple motions, we use a redescending M-estimator to compute rotation parameters of a dominant motion within the patch. To reduce computational and storage burdens of standard multidimensional Hough transform, we use adaptive Hough transform to iteratively refine the relevant parameter space in a "coarse-to-fine" fashion. Our method can robustly recover 3D motion parameters, reject outliers of the flow estimates, and deal with multiple moving objects present in the scene. Applications of the proposed method to both synthetic and real image sequences are demonstrated with promising results.

We evaluated six algorithms for computing egomotion from image velocities. We established benchmarks for quantifying bias and sensitivity to noise, and for quantifying the convergence properties of those algorithms that require numerical search. Our simulation results reveal some interesting and surprising results. First, it is often written in the literature that the egomotion problem is difficult because translation (e.g., along the X-axis) and rotation (e.g., about the Y-axis) produce similar image velocities. We found, to the contrary, that the bias and sensitivity of our six algorithms are totally invariant with respect to the axis of rotation. Second, it is also believed by some that fixating helps to make the egomotion problem easier: We found, to the contrary, that fixating does not help when the noise is independent of the image velocities. Fixation does help if the noise is proportional to speed, but this is only for the trivial reason that the speeds are slower under fixation. Third, it is widely believed that increasing the field of view will yield better performance. We found, to the contrary, that this is not necessarily true.

Presents a method to determine the 3D motion of multiple objects from two perspective views. In our method, segmentation is determined based on a 3D rigidity constraint. We divide the input image into overlapping patches, and for each sample of the translation parameter space, we compute the rotation parameters of patches using a least-squares fit. Every patch votes for a sample in the translation and rotation parameter space. For a patch containing multiple motions, we use an M-estimator to compute rotation parameters of a dominant motion. We use the adaptive Hough transform to refine the relevant parameter space in a "coarse-to-fine" fashion. Applications of the proposed method to both synthetic and real images are demonstrated with promising results.>

Applies mean field technique and presents a deterministic algorithm to determine the optical flow and motion boundaries. To deal with the problem of large motion, the authors present an adaptive multigrid approach, which also greatly reduces the computation time. This algorithm is fully parallelizable and iterative. The performance of the proposed method is compared against known algorithms using performance measures proposed by Barren et al. [1992]. Experimental results indicate that the present approach provides good estimates of optical flow and motion boundaries.>

Presents a robust method to determine 3D motion and structure of multiple objects. Rather than segmenting the scene containing multiple motions using 3D parametric model, the authors use the general 3D motion model and exploit Hough transform and robust estimation techniques to determine motion and segmentation simultaneously for an arbitrary scene. The authors divide the input image into patches, and for each sample of the translation space and each patch, the authors compute the rotation parameters using weighted least-squares fit. Each patch votes for a sample in the five-dimensional parameter space (translation and rotation). The multiple local maxima in the parameter space naturally correspond to the multiple moving objects. The authors' experimental results show that the proposed method is robust and relatively insensitive to noise.