Image Stitching using Watersheds and Graph Cuts

Image stitching is commonly used in many different applications. One such application could be creating panoramas or high resolution images. To create seamless images is often very computationally demanding. On the other hand, the technological evolution of the latest years has created a customer demand for these algorithms to be used in handheld devices, such as cameras and cell phones. The methods need therefore to be faster in order to work in near real-time on mobile devices. Using one of the simpler algorithms often results in degrading artifacts, such as artificial seams or blurred regions. To favor seamless blending and clarity in the blending areas, while keeping a low computational cost, is the main motivation for this report. In this report the implementation of a fast image blending algorithm, based on the watershed algorithm and graph cuts, is presented [5]. The images are first aligned using SURF-matches.