| With the rapid development of vision technology,its applications in the fields of medicine and bionics have become more and more extensive.Single-camera images are far from being able to meet the observation needs,so video stitching technology has become an important means to achieve large field of view observations.Based on the above requirements,this paper has carried out the research of four-channel video stitching technology and designed a set of FPGA-based video stitching system.This article first determines the main realization process of video stitching,and conducts in-depth research on the realization of the main steps,including image preprocessing,image registration,and image splicing.At the same time,this article optimizes the deficiencies and problems of the existing algorithms to meet the needs of the hardware platform.This paper proposes an optimized brightness correction method in the image preprocessing stage.This method introduces the concept of logarithmic entropy into the brightness equalization process,and realizes the global optimal solution through the process of logarithmizing the pixel brightness value.This method improves image brightness while eliminating vignetting.In this paper,an optimized ORB algorithm is used in the image registration step.This article first constructs the region of interest to remove unnecessary calculations and increase the calculation speed.Then the scale space is constructed to solve the problem that the traditional ORB algorithm has no scale invariance.Finally,the grid method is used to divide the image and combined with the quadtree algorithm and the non-maximum suppression algorithm to screen the feature points,to achieve a uniform distribution of the feature points,which significantly improves the matching accuracy.This paper proposes an improved optimal stitching algorithm to realize video stitching.The algorithm influences the search results of the stitching line by adjusting the weight of the pixel gradient difference in the calculation,so that the best stitching line position falls as far as possible in the background area without edge points.This method significantly reduces the visibility of the suture.In addition,this article also optimizes the suture search process: firstly search for sutures by down-sampling,then interpolate to obtain the initial sutures,and then adjust within a certain width neighborhood to obtain the precise best sutures.Compared with the traditional method,this method greatly reduces the amount of calculation and significantly improves the calculation efficiency.Finally,this paper implements a four-channel video stitching hardware platform based on FPGA chip design,and conducts video stitching experiments,and obtains better experimental results. |
