Segmentation of high resolution remote sensing images by combining hidden Markov random field model and fuzzy c-means at the region level

Abstract

Abstract. In high spatial resolution remote-sensing images, complex landscapes are usually accompanied with macro texture patterns, which often adversely affect segmentation accuracy, mainly due to their high spatial and spectral heterogeneity. To address this problem, this study develops an image segmentation method by combining the iteration procedure of fuzzy c-means (FCM) clustering and hidden Markov random field (HMRF) model at the region level. The performance of the proposed method was assessed through aerial images. Results indicate that the proposed method can improve image segmentation accuracy, compared to FLICM, HMRF-FCM, MRR-MRF, and IRGS.

Adaptive image segmentation method based on the fuzzy c‐means with spatial information

“…Hence, it has been successfully applied in the segmentation of general images [3,4], magnetic resonance [5,6] and medical [7,8] images. Many improvements have been made to FCM since its advent, such as the two-dimensional adaptive fuzzy c-means [9], the adaptively regularised kernel-based fuzzy c-means (ARKFCM) [10], and the Markov random field model combined FCM [11,12].…”