Improved Video Mosaic Construction by Selecting a Suitable Subset of Video Images

By stitching together adjacent images from a video sequence surveying a scene, a video mosaic of the entire panorama can be formed. Since a video survey consists of a sequence of images having small relative displacements with respect to each other, there is redundant overlapping information in consecutive images so that not all consecutive video images are required to create a mosaic, and only a subset of suitable images needs to be chosen. Images that are misaligned due to subpixel translation, rotation or shear are difficult to perfectly re-align and stitching of such images can result in a mosaic where discontinuities are noticeable. We propose a new technique for the construction of a seamless mosaic to minimise discontinuities. Our technique partitions an image into four quadrants to register with those of successive images. These registration values are used to form a misalignment index for selecting the best images for stitching.