Research Of Image Mosaic Basde On The Imaging Microscope

In the field of medicine and biology, microscope is an important tool for observation.The view of microscope image is reducing as the magnification increasing. In the low magnification lens, there is a wild perspective but no details. On the contrary, in the high magnification lens, the details can be seen clearly, but we can't get the whole sample. So, to get a image including clear details and wild perspective with microscope is an important thing . Image stiching technology can be used to achieve the target.In this paper, the cell images obtained by microscope are test images. There is much dot structure in the image, such as nuclear, mitochondria, membrane attachments. Image mosaic based on SURF is applied to this paper.The test images in this paper are 5 cell partial images. In the image preprocessing stage,we mainly solve the problem of noise. Firstly, introduce some common filter and make filtering process to the images. Then, detect the edge of cell image. According to the result of experiment, this paper chooses Gaussian filter to eliminate noise. This paper uses Canny edge detection to detect the edge of test images.In the image registration stage, firstly this paper useing Corner Detection Algorithm to detect the feature points, including Moravec Corner Detector, Harris Corner Detector and SUSAN Corner Detector. Then, this paper introduces SURF feature detection algorithm and detects feature points by this way. After experiment, the results are compared with Corner detection algorithm. According to the comparing result, SURF is the most accurate method for microscopy images. This method uses Fast Hessian Matrix to detect the feature points in different Scale Space, then establishes description vector for every feature point. After the vector established, the method calculates the Euclidean distance between different vectors to initially identify the feature points. In the processing of feature points matching, the false matching points are obtained around the feature points. At last, we use RANSAC to remove the false matching points to obtain the correct matching points.Finally, we splice and merge the test images based on microcope, and complete the task of frame to frame and frame to mosaic subgraph. The experiments show that SURF Feature Detection can effectively complete the task of microcophy images stitching.