An extraction unit extracts, as a candidate pixel of a disturbance object, a predetermined number of pixels in order of smaller distance from each distance image acquired by a distance measurement sensor, the predetermined number being one or more. A calculation unit calculates a feature value indicating a characteristic of temporal change between each candidate pixel in a distance image of a current frame and a corresponding candidate pixel in a distance image of a past frame. A removal unit specifies, as a pixel indicating a disturbance object, a candidate pixel the feature value of which calculated by the calculation unit is larger than a predetermined reference feature value, and removes the specified pixel from the distance image of the current frame.