Recovering Shape and Size from a Single Endoscope Image using Optimization

Abstract

Recovering the shape and size of a polyp from endoscope images is an important research topic in medical diagnosis. As a method for recovering 3D shape of polyps, the authors proposed a method that can recover the shape accurately and faster than the conventional method by deriving an equation based on the Eikonal equation from the condition that the optical and geometric constraint equations become equal from the reason that both equations determine the depth Z of the same point. To obtain the absolute size of polyp using this method, it is necessary to estimate the reflectance factor C of the object surface after converting the original image into the uniform Lambertian image. The reflectance factor is obtained by optimizing the error between the width of the entire image and the estimated width of the recovered result. The width of the entire image is defined from the assumed width of the blood vessel. The surface reflectance factor is estimated from a single endoscope image by introducing an optimization approach.