Implementation of depth estimation algorithm for smartphone camera

Abstract

Car navigation application is a raising topic in recent years due to rapid development of technology. Although traditional navigation application embedded into car system is working well, ones running on handheld devices such as smartphone or tablet is going strong and gradually have its market because of its portability and availability. Smartphone nowadays have many functionalities from which driver assisted system can utilize, but knowledge about relative position of the car to environment has become the missing piece of whole system. This project aim to implement a technique, which can run on commercial smartphone itself, extracting depth information from image captured by the phone’s camera. This work develops on basis of stereo imaging theory and combine various available image transformation technique. Although the technique used in this project is similar to ones that are deployed on common stereo camera system, rectification process is embedded with feature detection as well as feature matching algorithm to work with arbitrary image. The modified technique has been proved to work well under some conditions, although the robustness was still a concern.