An integer programming approach to hand pose estimation using depth data

This thesis introduces a novel approach to tackle the problem of estimating the pose of an hand from a single depth frame. We will come up with a scheme for finger and palm regions extraction, followed by a Mixed Integer Programming formulation to complete the process of finger segmentation. A key idea is a novel method which is able to rearrange the hand point cloud, through a simple transformation, in a way that better fits for recognizing its structure