Shape Recovery and Polyp Recognition from Endoscope Image for Supporting Medical Diagnosis

点击翻译按钮获取中文摘要

The purpose of the research is 3-D shape recovery of an object from endoscope image and new approaches were developed including the generation of diffused reflectance image, automatic detection of polyp candidate, and shape recovery under the condition of point light source illumination and perspective projection. The algorithm recovers the shape based on both of photometric and geometric constraints and it is assumed that the objective target includes convex and concave curved surface. To generate diffused reflectance image, developed approach absorbs the effect of difference of the reflectance parameters among the segmented regions in normalized color space.Automatic polyp detection approaches with high accuracy was also developed by extracting the polyp candidate region by Hessian filter and SVM learning with feature selections.