A Novel Seam Finding Method Using Downscaling and Cost for Image Stitching

Abstract

Seaming finding is an important step for creating panorama images because it smoothes away differences observed at boundaries between stitched images. We propose an improved seam finding method in which we define a cost function to measure the discrepancies that boundary pixels cause. We are also able to improve computing complexity by avoiding finding a seam over the whole area of overlapped region. Instead, we use a downscaled version of overlapped area to approximate a seam and then interpolate the seam to the original region. From experiments to generate panorama images, we compare our method with three other existing seam finding algorithms and observe that our method is able to produce better quality panorama image than the existing methods, while the processing time is comparable to those of the others.