Image Segmentation Based on Ball Vector Machine
Wenhai Wu, Huaxian Pan, Yao Zheng an, Guojian Cheng · 2012
Owing to the large scale of multi-dimensional datasets in image processing, the standard Support Vector Machine (SVM) has a high time complexity in the training process for image segmentation. A new machine learning method, Ball Vector Machine (BVM) is used for image segmentation in order to reduce the training time in this paper. The experimental results show that BVM has a similar segmentation effect and noise immunity performance compared to standard SVM for image segmentation in the condition of corrupted and none-corrupted. However, BVM consumes significantly lesser training time than the standard SVM. BVM can greatly improve the overall performance of image segmentation.