Automatic stitching method for spherical panoramic video

Panoramic video has become a hot research topic in the field of virtual reality in recent years. Existing research suggests that panoramic video cannot automatically be stitched yet. In addition, stitching accuracy and time consumption are not satisfactory. After analyzing the relationship among the several coordinate systems involved in shooting and images, this research puts forward a manual method and steps for stitching a panoramic video. Then a matching point searching algorithm based on Harris corner detection, empirical position and HSV matching is posited. With this algorithm, matching points are searched and found automatically, and panoramic videos are automatically stitched. Comparison experiments were carried out. The test results proved that the described automatic stitching algorithm for panoramic videos performs better than the alternatives in both accuracy and time consumption.