Research On Image Stitching Technology Based On Optimal Seam

Image stitching technology aims to quickly generate high-definition seamless images with wide field of view,and the generated images will not appear distortion or ghost.In recent years,this technology has developed rapidly and is widely used in people’s production and life,such as panoramic map,virtual reality and intelligent driving.However,due to the limitation of the equipment itself and the existence of registration error,the spliced images will have problems such as dislocation in different degrees and light color jump in the overlapping area,resulting in parallax between images.Therefore,how to ensure that the spliced images are clear,natural and ghosting free is an urgent problem to be solved in image splicing technology.In addition,in the presence of moving objects,how to improve the stitching quality and better meet the visual needs of human eyes is another problem to be solved in image stitching technology.Aiming at the problem that image stitching is prone to blur and ghosting in the scene with large parallax,an optimal seam algorithm based on mixed norm is proposed.When constructing the energy function of the optimal suture line,the algorithm introduces the alignment difference and uses the exponential form of alignment difference and color difference to strengthen the distinction between alignment and unalignment in the overlapping area.In this paper,comparative experiments of different suture algorithms are carried out.The results show that the optimal seam algorithms found by this method can successfully avoid passing through the misaligned area,making the stitched image look more natural and no ghosting.Aiming at the problems of foreground fracture and ghosting in image stitching in the scene with moving objects,a fusion algorithm based on moving object mask image is proposed.Firstly,the mask image of moving object is extracted by vibe algorithm,and then the unnecessary edge and noise in the mask image are removed by morphological processing.Finally,the mask image of moving object is applied to image fusion.In this paper,comparative experiments of different fusion algorithms are carried out.The experimental results show that this method can effectively improve the stitching quality of image in the case of moving object.Aiming at the time-consuming problem of image stitching algorithm,CUDA was used to accelerate and optimize the algorithm during image fusion.Meanwhile,the speed of fusion realized on CPU and GPU was compared,which finally verified the real-time performance of the algorithm in this paper.