Adaptive Nonlinear Diffusion Filter for Capsule Endoscopy Images

Different diseases in small intestine can be prevented and cured through early detection. The traditional detection methods such as endoscopy, ultrasound, and CT scan suffer from different shortcomings such as invasiveness, unclearness, and so on. The capsule endoscopy invented in 2001 can view the entire small intestine without pain, sedation, or air insufflation, so it has been widely used to detect the status of the small intestine. However, the images produced by the capsule endoscopy are often rather dark for many reasons and the resolution of them is only 256*256 due to the volume constraint of this little capsule. In order to facilitate the doctor’s analysis and diagnosis, we propose an adaptive nonlinear diffusion filter to enhance this kind of images. Based on the eigenvalues of the image’s local Hessian matrix, we devise a local coherence to measure the anisotropy of the local region. Then, we put forward an adaptive nonlinear diffusion filter with a new conductance function to enhance such images. Finally we report some experimental results on real images confirming the effectiveness of this new filter.