A Defocus Technique based Depth from Lens Translation using Sequential SVD Factorization

The imaging system has been processed by many researches because of supplying the variety of information in the environments. Recovering three-dimensional geometry of scene and motion of the camera has been studied for the position measurement and recognition of objects due to estimating the three-dimensional geometry of scenes from the stream of two-dimensional images. The researches for depth estimation such as stereopsis, motion parallax and blurring phenomena are based on the camera system and optical phenomena. Among the researches, depth from lens translation in a specific case of shape from motion is used for a zooming effect that causes by translating the lens center along the optical axis of the lens [1-5]. This effect can be used for both a fixed camera and an unfixed camera because the camera zoom function is equal to the camera translation itself along the optical axis of the lens. Depth from lens translation using the sequential SVD factorization recovers both the shape of an object and its motion from a sequence of images using many images and tracking many feature points to obtain feature position information. The shape and motion can be obtained by using an orthogonal matrix derived from singular value decomposition (SVD). The method robustly processes the feature trajectory information using SVD, taking advantage of the linear algebraic properties of orthographic projection. The sequential SVD factorization not only solves a problem of the initialization in a disadvantage of Kaman filter, but also processes more rapid than a method using the Kaman filter [6,