Quality evaluation-based iterative seam estimation for image stitching

Seam-driven approaches have been proven effective for handling imperfect image series in image stitching. Generally, seam-driven approach utilizes seam-cutting to find a best stitching seam from one or finite alignment hypotheses based on a predefined seam quality metric. However, the quality metrics in these methods are defined in order to evaluate the average performance of the pixels along the seam. In many cases, the seam with the minimal cost is not necessarily optimal in terms of human perception. In this paper, we propose a novel iterative seam estimation method where the iteration procedure is guided by our quality evaluation for the pixels along the seam. First, we estimate a stitching seam via the conventional seam-cutting and introduce a hybrid quality evaluation to evaluate the pixels along the seam; the evaluated costs are then used to recalculate the difference map of the overlapping region and re-estimate a stitching seam. This evaluation–re-estimation procedure iterates until the current seam changes negligibly comparing with the previous seams. Experiments show that comparing with the conventional seam-cutting and other seam-driven methods, our method can produce results with less artifacts and overall is more visually pleasing.