Efficient 3D image warping for composing novel views

This paper presents an efficient inverse warping algorithm for generating novel views by combining multiple reference images taken from different viewpoints. The method proceeds in three steps. Firstly, the reference images are preprocessed for extracting edge pixels. Secondly, an inverse warping is performed to render the desired image from one primary reference image. By taking advantages of epipolar line features and depth discontinuities in reference images, the inverse warping can be efficiently applied by segments, to accelerate the rendering substantially. Finally, holes in the desired image are filled up through searching the corresponding points in other reference images. At this stage, two accelerating techniques are presented. By using the proposed algorithm we can navigate a virtual environment at interactive rate