Locality-Constrained Dictionary Learning Classification Method Of Wce Images

Wireless capsule endoscopy (WCE) is widely used in medical diagnosis which generates lots of images for each operation. For clinicians, analyzing these images is a time-consuming and laborious task. Therefore, an automatic image classification algorithm is proposed to help doctors quickly check the condition of specific organs such as stomach or small intestine in the digestive tract. In the preprocessing stage, we obtain the regions of interest from the images of cecum, pylorus and cardia followed by extracting the fused color and texture features. At the next stage, an effective organ classification method based on dictionary learning was proposed, the locality-constrained term was constructed using the profile matrix and locality information of atoms. The simulation results demonstrate that the LCDL algorithm exceeds some existing algorithms on the organ classification task.