A novel superpixel approach utilizing depth and temporal cues

Abstract

In this paper, a novel approach to identifying superpixels in the cluttered environment is proposed. In our proposed method, the temporal cue and depth maps obtained from depth sensors are combined with the popular method SLIC for superpixels using a new formulation of distance-minimizing clustering. Under cluttered environment, this proposed method can, compared with color-based approaches, better identify the contour of objects. Experiments have been carried out using a public dataset to compare our approach to other methods. The experimental results demonstrate that our approach outperforms other approaches.