Study On The Method Of Blood Microscopic Multi-field Image Stitching

In clinical tests,the completion of blood tests often relies on the microscope.The medical microscopic image has a clear field of view under the high magnification microscope and is easy to observe local details.However,the field of view is too small to observe the whole specimen.Under the low magnification microscope,the field of view is large but it is difficult to observe the details.Therefore,image stitching technology plays an increasingly important role in medical examination.This study is mainly aimed at the characteristics of blood microscopy images under the forty magnification microscope.A multi-view image mosaic method based on blood cell images is designed to ensure the stitching accuracy and stitching quality,and it also improves the stitching speed.Main tasks in this paper includes:1.Design a fast and stable feature detection algorithm which is suitable for the characteristics of blood microscopic images.Firstly,it is necessary to discuss and analyze the principle,characteristics and application of the currently widely used SIFT,SURF and Harris feature detection algorithms.Secondly,take the problem of dense arrangement of blood microscopic images and low contrast between cells and background into consideration,a local texture based SIFT feature detection algorithm is proposed.In the algorithm,64 fields are taken as the center of the feature points and the gray scale information is binary coded.Meanwhile its statistical information is used as a part of the feature descriptor.The method makes full use of the texture information carried by the cell itself,and enhances the stability of the characterization operator.2.An improved RANSAC algorithm is proposed to reduce the computational complexity as well as ensuring accurate calculation of the homography matrix.In this study,the final correct matching pair is determined by two observations,which makes the RANSAC algorithm out of the limit of the number of iterations.The method firstly ensures that the stitching effect has high readability.At the same time,the computational complexity is significantly reduced,and the computational efficiency of multi-view stitching of blood microscopic images is improved.3.Design a multi-angle experimental comparison.In the module of feature detection algorithm,the fuzzy images,the noised images,the geometric deformation images,the rotation images and the images at different scales are compared with the mainstream feature detection algorithm.The experimental results show that the feature descriptors obtained in this paper are more stable.In the blood cell images with dense cell arrangement,the stitching accuracy rate is 63.91% higher than that of the SIFT source algorithm.In the image fusion module,the algorithm stitching results are smoother and natural without obvious stitching seams.And the MSE,PSNR and SSIM are selected to evaluate the mosaic quality of the panoramic image.Finally,50 sets of images are tested for the computational efficiency,and the statistical results are obtained.It shows that the average efficiency of the algorithm is 42.37% higher than that of the SIFT algorithm.This study solves the common problems in the stitching of blood microscopic images,and provides a large field of view and high magnification for the examination of blood diseases.It is hopeful to be further applied to blood cell testing.