Design And Implementation Of Embedded Electron Microscopic Image Stitching System

Image stitching technology of medical images gained by electronic microscope is an important supplementary mean to clinical diagnosis,treatment,and preoperative plan.And it is used to solve the difficult problem that image is reduced in vision when watched by electronic microscope with high magnification so that the observer can’t get the information of the integrated image.Currently,many image stitching algorithms are implemented on the computer by using software,with the development of electronic technology,Embedded System has features such as streamlined system,facing specific applications and high efficiency in developing,so it has obvious advantages and broad prospects when it is applied to develop medical equipment.In this thesis,based on the rich relevant domestic and overseas document,the careful and deep research has been done on image stitching.This thesis designed and realized a cartilage tissue slicing image stitching system based on Embedded System.The system designed a good man-machine interface,and includes five parts:the simple image acquisition auxiliary module,the image library building module,the image transmitting module,the image stitching module and the image similarity detecting module.Based on the comparison of performance of feature point detecting algorithm based on SIFT(Scale Invariant Feature Transform)and feature point detecting algorithm based on SURF(Speed Up Robust Feature),and in-depth research and trial and error of SURF,the thesis eventually used improved SURF feature points algorithm,the improved method is defining characteristic points detecting area.During the process of detection,the thesis used multithreaded approach based on the independence of the images to speed up the detecting process,and then the thesis adopted the BBF algorithm to search and match the SURF feature points quickly.At last,a simple method based on the characteristics of electron microscopic images is used for image matching.This thesis completed algorithms implementation and simulation testing of the system by using mixed programming C and C++ language and in the Qt platform under Linux operating system firstly.After that,this thesis selected ICETEK-AM3 517-KB-EZ evaluation as the hardware platform and completed algorithms transplantation and hardware implementation.System testing showed that the modules of the system can operate normally and can splice images that have overlapping area of 1/4 to 1/2 with high speed and good quality.Such as,when splicing two images whose overlap area is 1/2 and their size is 320x240,it can detect 48 pairs of matching points and cost 1:44s;when splicing two images whose size is 1280x960,it can detect 36 pairs of matching points and cost 3.76s;when splicing six images that have different overlap areas and whose size is 320×240,it costs 20.58s,and the similarity to the original can reach above 90%.The system is capable to stitch images independently without the computer so that it can be used as a dedicated matching equipment of the electron microscope.The system function is integrated,the device is convenient to carry and the code has very strong portability.So the system has practicability and exploitability in some extent.And it prepared basic researching material for developing 3D reconstruction equipment in the future.