Research On Pixel-level Image Fusion Method For Vehicle Application

With the rapid increase of car ownership,due to the driver's visual blind,the problem of traffic accidents is becoming more and more serious,the emergence of the vehicle image system based on image mosaic technology,to a certain extent,which eliminated the driver's visual blind spots,has gradually become the research focus in the field of intelligent transportation and image processing.As one of the key link of image mosaic process,the performance of the image fusion technology has a crucial effect on the quality of the fused images and visual effect.Therefore,the pixel-level fusion method based on vehicle image system has been conducted in deep research,the main research content as follows:Firstly,for finding splicing seam with low accuracy and can't eliminate splicing seam effectively in the image splicing,a splicing seam elimination method based on neighborhood comparison is proposed.First of all,use a 3*3 sliding template to find a pixel point whose gray standard deviation value is the largest in each row of number of rows continuously from the image overlap areas line by line,exclude the pixel points with large deviation until after two points in iteration,and calculate the mean value of the rest of points as the starting splice point of the splicing seam,set the threshold value to limit the range of candidate splicing point in turn on the basis of their column position.Finally,connect the optimal splicing point of each row as the optimal splicing seam.Experimental results show that the method can make the gray difference smaller on both sides of the splicing seam in order to eliminate the splicing seam relatively effectively,make the image overlap region smooth and natural,and realize seamless splicing better.Secondly,for producing splicing seam in the image mosaic,through the analysis of the hard correction method,a splicing seam elimination method based on correction ratio of gray average difference is proposed.First of all,calculate the gray average of the image on both sides of the splicing seam respectively,the ratio of the absolute value of gray average difference and original average,and get the correction ratio of each column on both sides of the splicing seam according to the ratio within the scope of the correction.Then multiply the original pixel gray value and the corresponding correction ratio of each column together to get the pixel gray value after correction.Finally,the corrected image is weighted fusion with the original image.The experimental results show that the method can eliminate the phenomenon of the splicing seam effectively due to the large pixel graydifference,the guarantee joining together information,clarity,and overall quality of the image at the same time,the efficiency is higher.Thirdly,aiming at the problem of the seam or ghost due to the different exposure or moving objects during image stitching,an image fusion algorithm based on the optimal seam and improved gradual fusion method is proposed.Firstly,in the overlap areas of two stitching images,the optimal seam is obtained according to the idea of dynamic programming,then the images are stitched along the optimal seam.Finally,the pixel gray values of fusion image are determined by fusion rules of improved gradual fusion method.Experimental results show that the method can make the image overlap areas more smooth and natural,and eliminate the phenomenon of ghost and stitching traces.In order to verify the performance implementation of the image fusion algorithm in vehicle panoramic system,firstly,a experiment platform based on combining the Direct show technology and telephone camera is build to simulate vehicle system,then the studied image fusion algorithm was applied to the system to verify the effect of image fusion.All in all,the studied image fusion algorithm based on the pixel-level,not only can effectively eliminate the seam of the image that has obvious gray difference,and realize uniform transition,but also is able to keep the information of target image for the images with moving targets,eliminate the phenomenon of ghosting around the target,finish the fusion work well,and provide the technical foundation to develop the practical vehicle image system.