Visual navigation can handle complicated problems, such as kidnapping, shadowing and slipping. A low-cost video camera is particularly suitable for mobile home robots in the sense of human robot interaction, and it does not disparity map computation. An efficient vision-based simultaneous localization and map building (SLAM) method is presented for home robots using a forward monocular camera. This paper also presents a novel framework of scale-invariant feature transform (SIFT), where the difference of Gaussian (DOG)- based scale-invariant feature transform method is replaced by the difference of wavelet (DOW) transform. The modified SIFT enables real-time applications or embedded systems for home robot products. Two different types of home robots, such as cleaning and service robots serve as a tested platform of the proposed vision-based navigation. The experimental results show that the robots can provide acceptable navigation performance on unstructured environment in real-time.