Shape from Color Consistency Using Node Cut

In this paper, we present a method to estimate the shape of a three dimensional (3D) object by enforcing color consistency in the views of object points in images acquired from different viewpoints. The estimated shape and color information of the object can be used to render new views of the object. We first construct a 3D voxel space and assign to each voxel a photo inconsistency value, and then represent this voxel space using a node capacitated graph. We use s-t node cut to obtain a surface that minimizes the summation of photo inconsistency values of surface voxels. Experimental results for simulated and real objects are provided. We also simulate how the calibration errors of camera intrinsic and extrinsic parameters affect the performance of our algorithm.