|
计算机科学 2006
A Novel Search-Transform Algorithm for Inverse Image Warping
|
Abstract:
As the basic algorithm of image-based rendering,inverse image warping is blocked to wide application for its high complexity and low performance,which is due to the complicate search-and-match process in reference field.In this paper,a novel search-transform algorithm (STA)is proposed to change the situation.On the observation that search-and-march process in desired filed is much simpler than reference one,STA transform most of this process from reference field to desired one based on the epipolar geometry characters.Meanwhile,STA takes full advantage of the spatial continuity of scene to effectively reduce the step length of searching by setting the self-adaptive warping order for desired pixels.STA is proved to be spare but efficient by experiments,which can eliminate most of the redundant computation and obviously accelerate the traditional inverse image warping algorithm.