A Study Of Microscopic Images’ Stitching Algorithm

Microscopic image stitching technology can be used to splice multiple small field images into a microscopic image of larger field of view, which can be widely used in medical image analysis. Based on the microscopic research on image mosaic technology at home and abroad, this paper focuses on researching image stitching algorithm based on gray-scale information and improves template matching algorithm and gradual fade-out fusion algorithm. The reconstructed image has no obvious stitching trace and the speed and accuracy for matching images are improved significantly. This article mainly aims at the three modules including in pretreatment, matching and fusion of microscopic image stitching technology. Specifically as follows:(1)Four methods including component method, average method, maximum method and the weighted average method are used for gray processing in the collected RGB microscopic images. The weighted average method is chosen in image graying through contrast and analysis. And then five kinds of edge detection method including Roberts, Prewitt, Sobel, Laplacian and LOG are used to detect gray-scale images of the microscopic edge. Considering the effectiveness and speed of detection, Sobel edge detection method is selected as the edge detection method for gray microscopic images.(2)Grayscale template matching algorithm is improved in this paper because the two methods which including template matching algorithm for binary images and template matching algorithm for grayscale images are not strong in Engineering applicability. The improved template matching algorithm based on the grayscale treat the focus of level midline of the template image and vertical midline in the area to be matched as searching starting point. Unsymmetrical cross search is carried out with 4 times the width of the vertical direction of the template image as the initial step length (horizontal initial step length is half of the initial step vertical). And then set the maximum point of the evaluation function value in asymmetric cross searching as the center point of the next search and the searching step size is halved. It will not stop iterating until the search is reduced to 1. The maximum value obtianed from evaluation function is compared with the threshold value, then we can successful according to comparing the maximum value obtained from evaluation function with the threshold value.(3)In order to improve the accuracy of image matching, the multi template matching has been used, and the template center is connected with the regional center and the matching point is judged by comparing the two line angle. Compared to the traditional grayscale template matching algorithm, the matching speed is improved and the matching accuracy is as high as 97.8%, and the stability of the algorithm is improved greatly.(4)Gradual fade-out fusion algorithm based on gray values is improved in this paper. Treat match point for the centerline, to divide integration region into two pairs approximately. Used the right image for a reference, The left pixels are compared with 4 neighborhood average of the right pixels. Brightness can be adjusted according to the difference and be fused on the basis of brightness adjustment. A smooth transition for matching microscopic images can be achieved after improving the gradual fade-out fusion algorithm based on gray values.(5)The microscopic image stitching simulation interface is designed in the MFC framework, then the improved microscopic image stitching algorithm is loaded into the man-machine interface. The interface is easy to operate and the parameters can be changed with need to achieve rapid microscopic image stitching.