Quadrotors are well - known unmanned aerial vehicle structures that have some advantages such as hovering, vertical take – off and landing, and low – speed flight. On the other hand, quadrotors are subjected to modeling and sensor uncertainties that lead to erroneous state estimation. Kalman filter has been proven to be the optimal estimator for the Gaussian distributed noise for linear processes. However, linear dynamical models of the quadrotors are not accurate representations of the systems due to nonlinearities, and coupling between the states. Extended Kalman filter (EKF) is proposed to solve the above issue. But, first order Taylor series approximation for the nonlinear state model may lead inefficiencies. For this reason, another Kalman filter framework is proposed that employs unscented transformation (UT). Unscented Kalman filter (UKF), can model the state distribution as Gaussian random variable to the third degree for arbitrary nonlinearities. So, in this study, unscented Kalman filter based estimation scheme is presented to overcome the sensor and model noises for nonlinear quadrotor attitude dynamics. According to the statistical analysis, the approach can estimate and reduce the mean absolute error, root mean square error and also variance of the noise for all attitude states.