This paper addresses the problem of quadrotor attitude estimation when only IMU sensor data are available. First, a Smooth Sliding Mode (SSM) algorithm is considered to provide an estimation of roll/pitch angles. In a second step, the estimates provided by SSM algorithm are used to reformulate the innovation term of the complementary filter in order to enhance the estimation of attitude angles. In addition, it is shown how the proposed solution can be able to reconstruct a bias in the gyroscope measurements. Simulation results obtained on the nonlinear model of quadrotor illustrate the benefits of the proposed scheme.