Image segmentation by correlation adaptive weighted regression

Image segmentation aims to partition an image into several disjoint regions with each region corresponding to a visual meaningful object. It is a fundamental problem in image processing and computer vision. Recently, subspace clustering methods shows great potential in image segmentation. In this work we formulate image segmentation as subspace clustering of image feature vectors. To extend the capture ability of image varieties, we use a union of three kinds of feature including CH, LBP, and HOG. We propose an explicit data-correlation-adaptive penalty on the representation coefficients by a combination of correlation weighted ll-norm and 12-norm, and formulate the subspace representation as a Correlation Adaptive Weighted Regression (CAWR) problem. It can be regarded as a method which interpolates SSC and LSR adaptively depending on the correlation among data samples. It has subspace selection ability for uncorrelated data as well as grouping ability for highly correlated data. Experimental results of image segmentation show that the proposed model is better than the-state-of-art methods. (C) 2017 Elsevier B.V. All rights reserved.