Research On Local Adaptive Panoramic Image Stitching

As an important research area in computer vision,image stitching has been extensively and intensively studied during the past decades.Given a set of input images,image stitching aims at generating a high-quality panorama with the lowest computational cost.The key issue is the handling of parallax.To address this problem,numerous stitching approaches have been proposed,among which the local adaptive methods could achieve better results than traditional ones.This paper focus on local adaptive image stitching and propose several approaches which could achieve both effective and efficient image stitching.The main work and innovative points are included as follows:(1)Image stitching based on fast structural deformation along the seams.For a group of input images,first,the existing double-seam selection scheme is improved to more effectively search two distinct but structurally corresponding seams in the two original images.Second,along the double seams,one dimensional feature detection and matching is performed to capture the structural relationship between the two input images.Third,after feature matching,an efficient algorithm of linearly propagating the deformation vectors is proposed to eliminate structure misalignment.At last,image intensity misalignment is corrected by rapid gradient fusion based on the successive over relaxation iteration(SORI)solver.A principled solution to the initialization of the SORI significantly reduces the number of iterations required.Experimental results show that the proposed method outperforms the existing ones compared in terms of overall stitching quality and computational efficiency.(2)Image stitching based on local facet approximation.First,by introducing the camera motion model,a new planar transform model is proposed which is more stable than existing ones.Second,the scene is approximated as a combination of neighboring facets,such that the local adaptive stitching field is constructed using a series of linear systems about the facet parameters,which enables the parallax handling in three-dimensional space.At last,based on the local facet model,three facet approximation strategies are proposed including two weighting approaches and a triangulation approach.Moreover,the planar and spherical triangulation techniques are introduced to stitch perspective images and fish-eye images respectively.The efficiency of the proposed methods are verified through the comparative experiments on several challenging cases both qualitatively and quantitatively.(3)Image stitching method based on robust elastic warping.Given a group of point matches between images,first,an analytical warping function is constructed to eliminate the parallax errors.Then the input images are warped according to the computed deformations over the meshed image plane.At last,the seamless panorama is composed by directly re-projecting the warped images.As an important complement to the proposed method,a Bayesian model of feature refinement is proposed to adaptively remove the incorrect local matches.This ensures a more robust alignment than existing approaches.Moreover,our warp is highly compatible with different transformation types.A flexible strategy of combining it with the global similarity transformation is provided.The experimental results demonstrated that the proposed approach could achieve accurate alignment and efficient image stitching simultaneously.