An image processing system, which estimates a position and a pose of a camera and performs three-dimensional reconstruction processing of an object having a cylindrical shape by using a photographed image acquired by photographing an inside of the object with the camera, includes a processor. The processor estimates a first center axis of the cylindrical shape by using a group of three-dimensional points obtained by reconstructing, in a three-dimensional space, a group of feature points extracted from the photographed image. In addition, the processor performs bundle adjustment for correcting the position and the pose of the camera and the coordinates of the group of three-dimensional points by using a condition for minimizing a total sum of differences between a radius of the cylindrical shape and respective distances from the first center axis to individual three-dimensional points that constitute the group of three-dimensional points.