Research On Fast Video Splicing With Trapezoidal Distortion Correction Based On Harris Algorithm

In recent years,how to use image processing technology to splice video images with overlapping areas,making it a large scene video image with more image information is a hot research topic in the field of image processing.At the same time,for cameras,especially surveillance cameras,the trapezoidal distortion of the camera plays an important role in the effectiveness of the images taken by the camera.Therefore,there is an in-depth study on the fast video stitching technique of trapezoidal distortion correction based on Harris algorithm.Aiming at the disadvantage that the general trapezoidal distortion correction algorithm is difficult to transplant,a two-step projection method is used to correct image trapezoidal distortion.For the camera deflection angle is difficult to measure,this essay proposes a pixel-based angle detection algorithm that combines the image pixel coordinate system with the spatial geometrical relationship.The deflection angle of the camera is determined by a software method.At the same time,it can be well measured in the case of changes in camera height and angle.Experiments show that the algorithm can effectively correct the trapezoidal distortion of stitched images.In order to further improve the performance of Harris corner detection algorithm and improve its shortcomings,such as performing feature point detection on each pixel in the image which leads to a slowly speed,difficult to applied in actual engineering application.By learning the Harris corner detection algorithm based on the difference of 8-neighborhood pixel,I optimize the algorithm of the two-step screening candidate corners for the comparison and analysis of the top,bottom,left,and right neighboring pixels first,and then perform the comparison and analysis of the remaining 4 pixels.The new algorithm can effectively filter out most of the pixels that do not meet the conditions before detecting the feature points of the image pixels and improve the efficiency of the algorithm.At the same time,compared with the Harris corner detection algorithm based on the 8-neighborhood pixel difference,the algorithm increases the number of matching pairs and improves the detection speed.When we use the original fade-in and fade-out weighted image fusion algorithm to fuse the aligned images which have moving objects in the overlapped area,there will be many problems,such as “ghosting”,after image fusion.In addition,the original fade-in and fade-out algorithm may appear an unnatural fusion effect at the image boundaries.Aiming at these problems,an improved fade-in and fade-out weighted image fusion algorithm is proposed.Experiments show that the improved algorithm can solve these problems.At the same time,it appears a more natural fusion effect at the coincident region image.