Wavelet-based seamless image stitching under l1 and l2 norm

Image stitching is used to combine several individual images having some overlap into a composite image. In this paper, a novel wavelet based seamless image stitching algorithm is introduced, where the dissimilarity is defined in the gradient and curvature domain. In order to reduce the artifacts in image stitching due to the misalignment of image registration, we design a new energy function combining different norm function for different frequency subbands of wavelet transform according to the priors of multiresolution image representation. The good stitched image will optimize the energy function by efficient convex optimization algorithm. The experimental results show that our method can produce visually pleasant results when trivial misalignment exists.