Automatic detection of colonoscopic anomalies using capsule endoscopy
In this paper, a new method is proposed to detect abnormal regions in colonoscopic images by patch-based classifier ensemble. Through supervised learning from image patches of various sizes, a set of basic SVM classifiers is trained for each size. A diagnostic model can then be constructed based on the ensemble of basic classifiers which is then used to detect abnormal regions in colonoscopic images. The multiple sizes of patches provide multiple level representation of the image content, which can help improve detection results. Several fusion criteria are explored to determine the final output of the ensemble. Experimental results show promising performance of our proposed method.