Rectification of a mobile multi-camera chain based on hardware pose measures

In this paper we present an image rectification method for a multi-view image set captured by a mobile and articulated camera arrangement. The cameras are located in individual modules, each module consist in a camera, a single actuator and a proprioceptive angle sensor (servomotor encoder). Each module is connected to other modules in a linear chain configuration, with the restriction that the optical axis of all cameras are coplanar. The rectification proposed is based on an initial calibration-rectification with a known pattern. Then, a continuous self-rectification is proposed by using the Kalman filtered measures of the chain actuators. The rectification technique proposed is tested with real measurements from an articulated arrangement, results shown a fast and reliable sub pixel accuracy in the rectification.