ALIC: A Superpixel Segmentation Algorithm Based on Autonomous Attachment

In this paper, we reconsider the problem of low computational efficiency in traditional superpixel segmentation methods based on clustering. We propose a superpixel segmentation method based on autonomous attachment named ALIC, which use a simpler and more efficient distance measurement method to accelerate the algorithm. Simultaneously, the allocation of pixel autonomous attached fully considers the natural continuity between pixels, and each pixel can share label with its neighbors, which allows us to obtain better boundary performance. In the experiment, our method only achieves convergence in five iterations. On the basis of obtaining more sensitive boundaries, our algorithm improves the operation speed. (On a simple CPU core, it only takes about 0.1s to segment a 481x321 image into 400 homogenic superpixels).