Study on the Method of Blood Microscopic Multi-field Image Stitching

Abstract

In the diagnosis of medical blood diseases, there are contra-dictions between the clear view and the size of view under the microscope. In order to obtain clear blood cell images under a large view, this paper proposes an image stitching method for multi-view blood microscopy images. The method firstly preprocesses the input image sequence, and then uses the SIFT feature and the local LBP feature to extract the feature points of the image sequence, obtains the matching point pairs according to the threshold method, and then uses the improved RANSAC algorithm to calculate the homography matrix between the images. Finally, the weighted average in image fusion is used to realize the seamless stitching of multiview images. The experimental results show that the improved feature detection algorithm has good performance in the rotary image, blurry image and distorted cell image. The improved RANSAC algorithm effectively improves the computational efficiency of the image, and finally achieves multi-view blood display with high efficiency and seamless stitching of micro images.