Multi-View Images Rectification based on Common Points Extraction

The viewers can enjoy three-dimensional experience with the free-view stereoscopic video systems. In addition, they can switch the viewpoints as they like. But the vertical disparity will affect the quality of synthetic stereo images and generate blurring stereoscopic images. The stereoscopic images synthesized with severe vertical disparity can not be accepted by viewers as if the objects “jump” into the vision. Therefore, multi-view images rectification is one of the key steps to improve the quality of 3DTV viewing. Most existing rectified technologies are for binocular stereo images and further research is needed in the field of multi-view images. Pre-calibration parameters are required in the rectification based on multi-camera calibration, which is not convenient. In this paper, we propose a rectification method for multi-view images to eliminate the vertical disparity without using calibration parameters. First, SIFT (scale-invariant feature transform) is bring in to extract and match features automatically. Then the algorithm makes use of the relationship between the matched points, selects the unique standard image plane to calculate the rectification transformation matrix of each view images. The proposed method reduces the vertical disparity effectively, limits the rectified error propagation, so that to maintain the quality consistency for the rectified images shown in the experimental results.