G-Image Segmentation: Similarity-Preserving Fuzzy <i>C</i>-Means With Spatial Information Constraint in Wavelet Space

G-images refer to image data defined on irregular graph domains. This work\nelaborates a similarity-preserving Fuzzy C-Means (FCM) algorithm for G-image\nsegmentation and aims to develop techniques and tools for segmenting G-images.\nTo preserve the membership similarity between an arbitrary image pixel and its\nneighbors, a Kullback-Leibler divergence term on membership partition is\nintroduced as a part of FCM. As a result, similarity-preserving FCM is\ndeveloped by considering spatial information of image pixels for its robustness\nenhancement. Due to superior characteristics of a wavelet space, the proposed\nFCM is performed in this space rather than Euclidean one used in conventional\nFCM to secure its high robustness. Experiments on synthetic and real-world\nG-images demonstrate that it indeed achieves higher robustness and performance\nthan the state-of-the-art FCM algorithms. Moreover, it requires less\ncomputation than most of them.\n