Research On Image Stitching Techniques With Large Parallax

Image stitching or photo stitching is the process of combining multiple p graphic images with overlapping fields of view to produce a segmented panorama or high-resolution image. With the application of this technology is more and more widely, it becomes an active topic.The core modules of image stitching technology include image registration,image warping and image fusion. When the camera is stationary or the scenes lie in the same plane, the parallax between the images is small, and there are a lot of mature methods. But if shooting scenes lie in different planes, there will be differences in depth, resulting in large parallax. It will canse great obstacles to image registration. Consequently, the image stitching technique with large parallax is still a challenging task.In this thesis, we firstly introduce some classical algorithms and techniques in image registration, image warping and image fusion and then we evaluate and compare these algorithms. On this basis, aiming at the problem of image stitching with large parallax, we utilize the image warping technique based on content protection. We provide an improved image saliency detection algorithm, key regional protection strategies, to achieve better effect of stitching.If there is large parallax between the images, traditional image mosaic and fusion technology will lead to heavy shadow, blurred and other obvious flaws. In this thesis, we introduce the technique of stitching line based on graph cut, which can effectively reduce and eliminate the problem caused by parallax.