Research On Image Mosaicing Techniques Of Fundus Camera

With modern large-scale rapid development of medical imaging technology, image mosaicing technology has become a hot spot of the medical image processing and analysis. Digital fundus camera is a medical device to check for fundus disease, and can provide clear fundus images, help doctors make an accurate, objective and comprehensive analysis and decision. However, the imaging area of fundus camera is relatively small to the entire retina, while in the clinical research there are needs of the image information of the entire or most regions of retina, so we need to match multiple images by image mosaicing technology.This paper first introduces the development process and research status of image mosaicing technology, then in-depth analysis the two key technologies of image mosaicing: image registration and image fusion. The main researches of this paper are as follows:(1) For the characteristics of fundus images, it primarily studies the image registration techniques based on feature and area. The comparison of experiments shows that SUSAN corner detection and Harris corner detection method do not apply to fundus images registration.(2) It proposes an improved template matching method which is based on area. Grab initial matching points in pending registration images, and do template matching by the greatest similarity criteria to find the best matching point. This method has achieved fundus images registration well.(3) It studies fundus image fusion algorithms, and uses linear transition method in overlapping regions for registrated fundus images to eliminate splicing gap and achieve a seamless fusion results.(4) Take C++ Builder as a platform, complete image mosaicing module and implement seamless image mosaicing of any number of different field fundus images in "fundus camera software system". By practical applications, the image mosaicing method proposed in this paper has fast calculation speed, high precision and good robustness, and it has a strong practicality to fundus images.Image mosaicing technology facilitates diagnosis of various diseases and avoids many human errors. With its good reliability, stability, repeatability and accuracy, it has a wide range of research value and a very important clinical significance in medical image processing.The paper has constructive guidance for how to work out a suitable method of image mosaicing for fundus images.