CSA: An Accelerated Algorithm of Inverse Image Warping

Through seeking analytical solution instead of pixel searching , a novel inverse image warping method called CSA(cross-segment algorithm) is designed in this paper to accelerate the traditional method. By the new method, firstly,CSA marks the edge-pixels according to the depth information saved in the lists in special orders along the epipolar lines during warping. Then, edge-pixels in the primary reference field are mapped into the secondary one where the traces of neighboring edge-pixels are recorded. Combined with the epipolar line, if the location of the corresponding reference pixel can be calculated by solving their equations, the mapped pixel is then found. Otherwise, holes are marked. Finally, taking advantages of nearby pixels or other reference information, holes left are filled up. Since only edge-pixels rather than all the pixels in epipolar line are processed, CSA proves to be effective in accelerating the inverse image warping.