Classification of endoscopic images using delaunay triangulation-based edge features

In this work we present a method for an automated classification of endoscopic images according to the pit pattern classification scheme. Images taken during colonoscopy are transformed using an extended and rotation invariant version of the Local Binary Patterns operator (LBP). The result of the transforms is then used to extract polygons from the images. Based on these polygons we compute the regularity of the polygon positions by using the Delaunay triangulation and constructing histograms from the edge lengths of the Delaunay triangles. Using these histograms, the classification is carried out by employing the k-nearest-neighbors (k-NN) classifier in conjunction with the histogram intersection distance metric. While, compared to previously published results, the performance of the proposed approach is lower, the results achieved are yet promising and show that a pit pattern classification is feasible by using the proposed system.