In this paper, we propose a new approach to map-based indoor localization for walking people using an Inertial Measurement Unit (IMU). Generally, an office building includes various components such as corridors, rooms, floors, staircases, elevators. We consider the components for personal positioning. We present two mapping methods that are a map representation method and a position compensation method. The map representation method is applied to classification and representation of components in an indoor building map. The position compensation method is derived based on the distance and direction estimation using a dynamic scale factor. To evaluate our proposed methods, we have developed an indoor localization device using a small IMU and experimented on an office building. We show that our proposed approach achieves higher accuracy than traditional dead reckoning navigation.