Image Segmentation Using KFTBES

摘要

We present a supervised algorithm to improve the image segmentation algorithm based on texture and boundary encoding. Our method is due to the analysis of the implementation and result of the TBES algorithms, and we increase the adaptability of the TBES algorithms. Through constructing the train dataset with fine-class segmentation, our method adaptively distribute the optimum segmentation standard to each image using Kernel Fisher algorithms. We also compare our method to other similar popular algorithms and our method achieves the state-of-the-art results on Berkeley Segmentation Dataset.