A framework based on the Affine Invariant Regions for improving unsupervised image segmentation
This paper describes a novel fast mean shift algorithm based on a resampling technique with marked regular pyramid structure. This new method focuses on solving the problem of high calculation complexity when high data dimension or large data sets are involved in mean shift. By resampling the original image with marked regular pyramid structure, improved method reduces the number of pixels requiring mean-shift iterations and also reduces the complexity of the mean shift algorithm. The proposed approach is efficient in providing good segmentation performance. The experimental results demonstrate the effectiveness of the proposed approach.