Feature points based image registration between endoscope image and the CT image

Abstract

According to the special requirements of the minimally invasive surgical robot system, combined the CT images of the lesion site of the patient with the data implied in minimally invasive surgical robot, proposed a new fast feature point based local image registration method, can realize the registration for feature points on special tissue and general tissue. This algorithm introduces Euclidean distance between feature points as the measure of image registration; select the ICP algorithm framework to quickly solve the registration transformation function. Through testing of medical images, experimental results show that the algorithm can match the overall structure of the image information but also align the interested physiological anatomy position of image. Registration results are very satisfactory. This method is a fast, accurate and robust medical image registration method.