Study Of Panorama Display Technology Based On SIFT Feature Image Stitching

The 21 st century is an era of information globalization with an exponential increase in information volume.As an important form of information,images play an increasingly important role in the transmission of information nowdays.With the rapid development of economy and popularization of science and technology,the amount of information that people are receiving has increased exponentially,and the quality requirements for images have become increasingly high.Therefore,panoramas which have a wide viewing angle,high resolution,and a realistic sense of presence are getting more and more attention.Conventional digital cameras which are subject to their own hardware structure can't directly shoot a wide-angle and high-resolution panoramic image of a large scene.As a common image processing method,the image stitching technology can solve the above deficiencies by the post-stitch processing,and it has been extensively accepted in the processing and stitching of panorama in recent years.Although the image stitching technology is relatively mature nowadays,in the actual application process,due to the complexity of the scene environment,sometimes the use of traditional methods in some specific scenes can not achieve good image quality,and it can not meet the real-time requirements.Accoridng to the basic conditions described above,in this work,we elaborate and study the key steps in the image stitching technology,and present an efficient solution for panoramic display which is based on image stitching by improving and optimizing some existing technical bottlenecks.The specific contents are as following in four items:1.The basic theory of image stitching technology has been elaborated and studied.Through the implementation and analysis of image feature detection,match,and fusion of panoramic image generation methods,a set of high-efficiency image stitching programs was designed and good results were achieved.2.In the image preprocessing section,several common denoising algorithms are studied and compared,and an improved NLM algorithm is proposed by the author of this paper.By improving the original weighting method of the NLM algorithm,the noise processing capability is effectively improved,and the image quality is improved,thereby improving the efficiency of the following image registration and fusion.3.In the image preprocessing section,aiming at the phenomenon that the traditional cylindrical projection algorithm will produce image edge aliasing,an image edge aliasing elimination algorithm for cylindrical projection is proposed,which effectively eliminates the edge aliasing which was generated after the image cylinder projection,thereby improving the image quality.4.In the image registration process,an improved SIFT feature detection and matching algorithm is proposed for the shortcomings of the SIFT algorithm that require long operation time and low matching accuracy.Through the improvement of feature detection and matching methods,the operation time is effectively reduced,the matching accuracy is improved,and the quality of image registration is significantly improved.In summary,a large number of images were collected and stitched,and the results showed that the algorithm proposed in this work,which is based on SIFT feature image stitching,can effectively improve the computational efficiency and the quality of panoramic images.Thus,high-efficiency and high-quality panoramic display ultimately can be achieved.This work paves an effective way for the panaromic display with high efficiency and high image quality.