A new optimal seam finding method based on tensor analysis for automatic panorama construction

Optimal seam finding is a popular panorama construction strategy. It has the advantage of constructing panorama rapidly as well as being memory efficient. However, it is intrinsically sensitive to radiometric distortions, which is why artifacts are often visible in the composed panorama. Besides, it requires users to specify weight parameters, making such algorithms difficult to configure. In this paper, by fully exploiting the camera model and utilizing the knowledge of tensor analysis, we propose a novel optimal seam finding method. It not only enables the creation of artifact-free panorama from distorted input images but also requires no user-specified weights. We verify the effectiveness of our method by testing it on variegated sets of images with different types of optical distortions.

► We prove that some panorama construction methods are sensitive to distortions.
► Our method is robust against complex distortions like lens vignetting.
► Our method requires no user-specified weights.