View Synthesis via Wavelet-Based Morphing
小波空间的视图变形合成

This paper presents a new view synthesis technique using image morphing and 2-D discrete wavelet transform. The paper aims at pairwise images that are known without camera calibration and any other geometric information. Firstly, estimate the fundamental matrix related with any pair of images by techniques from stereovision. Using fundamental matrix, any pair of image planes can be rectified to be parallel and their corresponding points lie on the same scanline. Secondly, the paper describes an image morphing algorithm running in the wavelet domain. The proposed algorithm can synthesize new views more quickly. Any quantization techniques can be embedded here to compress the coefficients in depth. When displaying, compressed images are decoded and an inverse wavelet transform is achieved. Finally, a postwarping procedure is used to transform the interpolated views at desired position. Nice features of using wavelet transformation are its good locality in both space and frequency and its multiresolution representation mode, which makes the generated views be refined progressively and hence is suitable for communication. Also, the paper discusses the influences of different kinds of wavelet on the results.