52 Unsupervised Image Segmentation via $L1-Norm$ Minimization Learning

Unsupervised image segmentation (UIS) is a very challenging problem in computer vision and artificial intelligence. In general, the UIS can be viewed as a clustering problem, which aims to partition the image into K sets of regions that have coherent color and texture features. Clearly, the UIS results depend on the employed clustering methods. Inspired by the recent rapid progress of $l1-norm$ minimization techniques and the great success of $l1-norm$ minimization based sparse representation (SR), in this paper we propose a K-sparse clustering algorithm to segment the image into K partitions. The given image is first over-segmented into many small regions, from which the K clustering centers are learned under the SR framework using $l1-norm$ minimization. Finally, each of the regions can be classified into one of the K classes based on its SR coefficients over the K cluster centers. The experimental results on the benchmark images demonstrate the effectiveness and superiority of the proposed K-sparse clustering method to the classical K-means clustering method and recently developed sparse subspace clustering method in applications of UIS.