Abstract: There are various high dimensional engineering and scientific applications in communication, control, robotics, computer vision, biometrics, etc. Where researchers are facing problem to design the intelligent and robust system through high dimensional neural network. Although, some of the researcher used different structures of conventional neural classifier to solve the problem associated with high dimensional parameters. These network structures possess complex network, hence, very time consuming and weak to noise. These networks are also not able to learn magnitude and direction of each component simultaneously at 3D space. The quaternion is the number which possesses the magnitude and phase information in all four directions. This study presents the quaternionic domain neural network and its generalized learning algorithm that can finely process magnitude and phase information in respective directions. Its learning and generalization capability presented through different simulations demonstrate its applicability in 3D imaging.