A grouping-feature and Nesting-Kernel scene image segmentation algorithm
Shuqiong Xu, Cailian Zhu, Conggui Yuan · 2014
A framework of Grouping-Feature and Nesting-Kernel Support Vector Machine (GFNK-SVM) methodology is proposed to achieve a more reliable and robust segmentation performance. Firstly, the pixel wise intensity, gradient and SMF features are extracted to provide multiple features of the samples of GFNK-SVM model. A new clustering method called as Clustering Validity-Interval Type-2 Fuzzy C-Means (CV-IT2FCM) clustering algorithm is also presented to improve the robustness and reliability of clustering results by the iterative optimization. A type-2 fuzzy criterion is integrated to handle uncertainties in the clustering optimization process and CV is employed to select the training samples for the learning of the novel SVM model. Finally, by integrating SVM with a novel Nesting-Kernel, a systematic GFNK-SVM framework is presented and its model is trained as classifier for scene images segmentation. The GFNK-SVM scene segmentation method combined the advantages of multiple features and multiple kernels. Extensive experimental results on the BSDS dataset demonstrate that our method can obtain better performances than those state-of-the-art segmentation techniques.