An automated technique for high speed segmentation of endoscopic images

This paper introduces an automated technique for high speed segmentation of lumen regions from endoscopic images. The proposed technique is a two-step process. First, a novel adaptive image thresholding process, based on statistical and linear discriminant analysis, is employed to obtain a preliminary region of interest of the lumen region. Next, an iris filter is applied for enhancing the boundary of the region of interest to obtain an accurate detection of the lumen region. A comparison with existing techniques shows that the proposed technique is capable of high-speed and accurate segmentation.