Stitching Strip Determination for Optimal Seamline Search
Atik Garg, Lan-Rong Dung · 2020
To create a panorama image, various image stitching methods are available, which stitch two or more images together into a single image. However, often image stitching algorithms fail to stitch good panoramas. These failures in image stitching typically occur as the optimal seamline strategies in the image stitching framework are unable to avoid all the artifacts in their pathfinding. In our observation, working principles of optimal seamline strategies do not directly relate to the locating and avoiding artifacts. Thus, these principles often create errors in the construction of a panorama image. To overcome the issue, this paper introduces a new functional block, called Stitching Strip Determination, into existing image stitching framework. The new add-on block helps the existing optimal seamline strategies in identifying an artifact-free stitching strip/sub-region in the overlapping area which minimizes the possibility to incur artifacts in the image stitching process.