Unsupervised Moving Object Edge Segmentation Using Dual-Tree Complex Wavelet Transform

Abstract

Unsupervised moving object edge segmentation in dual-tree complex wavelet transform domain is presented. The interframe change detection method with automatic thresholds in six complex wavelet subbands is used with edge map detected by Canny edge detector in the corresponding frame in order to detect moving edges. We propose automatic threshold selection algorithm which are used for thresholding magnitude of subbands of dual-tree complex wavelet transform. The motivation behind using dualtree complex wavelet transform rather than discrete wavelet transform is the fact that the directionality information of dual-tree complex wavelet transform is better than the discrete wavelet transform and it is nearly shift-invariant. The performance of the proposed algorithm is demonstrated both numerically and visually.