1. A framework for robust motion estimation and segmentation in adverse outdoor conditions.
- Author
-
Rao, Sana
- Abstract
Robust motion estimation and segmentation are two important interrelated tasks with a variety of applications. Traditional motion estimation and segmentation methods focus on normal scenarios. However, they may fail to robustly estimate and segment the moving objects because of the ill-defined edges and boundaries in these scenarios. In this paper, a robust framework for motion estimation and segmentation in adverse outdoor conditions is presented. In the proposed framework, a non-local total variation motion estimation method (NLTV- L 1 ) with a support weight function is proposed. It is efficient to preserve the edges and handle the smoothness near the edges using a coarse-to-fine strategy for adverse outdoor scenarios. To deal with the segmentation problem, we take advantage of the motion pattern based background-foreground layer extraction to segment the moving objects. Such a way effectively groups the pixels that move in a similar direction and thus these pixels can be extracted from the noisy background. In addition, our framework does not require any post-processing step to remove noise. We demonstrate that the proposed framework outperforms other state-of-the-art methods on two commonly used datasets (i.e., Middlebury and MPI Sintel) for motion estimation and segmentation. Moreover, the experiments on real outdoor benchmarks (i.e., Foggy Zurich, Traffic and CDnet2014) show the robustness and efficiency of our framework in adverse outdoor conditions. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF