An automatic panoramic image mosaic method based on graph model

Image mosaic plays an important role in the fields of computer vision, robot navigation and virtual reality, and has become an active research field in recent years. There is a problem that the error will be accumulated and amplified in the case of aligning multi-images. This paper analyzes the looping path problem causing error accumulation, and introduces a multi-image stitching method based on graph model. We name the algorithm Weighted Shortest Path Algorithm, by which images can be stitched automatically. Matching Mean Square Error is introduced as the weight of edges on the graph, which is intuitive and easy to compute. Furthermore, the optimized Dijkstra algorithm is applied to speed up the path finding algorithm. Experiments show that the proposed algorithm causes less Matching Mean Square Error and obtains more stable results than other similar methods. Moreover, we extended our model to 360 A degrees panoramic image generation, and the quality of the stitched panorama is quite good.