Research On High-resolution Full-field Microscopic Imaging Methods

Microscopic imaging systems have been widely employed in material inspection and biomedical field.However,there are still some crucial problems that need to be solved such as inability to quickly perform automatic focusing operation and narrow fileds of view.In this thesis,automatic focusing and microscopic image stitching methods were investigated.The proposed solutions could provide an useful refrence for developing high-resolution full-field microscopic imaging systems.First,the automatic focusing methods of microscopes involve three aspects,including evaluation of image sharpness,selection of the position of the focus window,and strategy for searching the widow.Based on analyzing the performance of the typical image sharpness evaluation functions,a new function combining the Laplace and Roberts functions(L-R function)was emplyed to evaluate the image sharpness.The results show that all three indices of image sharpness evaluation functions could achieve their optimal values.For the selection of focusing window,to solve the problems of difficulty to capture to the main part of the target object and time-consuming,a method based on the scale-invariant feature transform(SIFT)algorithm was proposed to select the focusing window.The experimental results showed that this method has advantages over the central region window picking method in capturing the main part of the target object and over the multi-region method in computational complexity.In addition,this method is capble of picking the window position successfully in both cases of out-of-focus and in-focus images with a good adaptability.Second,microscopic image stitching methods mainly involve three aspects,including extraction of feature point,feature point matching,and image fusion.Due to some side effects similar to image rotation,translation,scale transformation,and noise,etc.during the image acquisition process,the commonly used feature point matching algorithms were compared.Then,the SIFT algorithm was investigated.Since in this algorithm the selection of matching points is based on Euclidean distance,it can yield low matching accuracy.In this context,the cosine similarity function was utilized to screen the matching point pairs,which led to a great decrease in number of matching point pairs.The experimental results showed that the accuracy of point matching was further improved after optimization and that the results of image stitching depend not only on the accuracy of matching points,but also on the approach for processing the pixels of the overlapping part of the two images.Therefore,a weighted average method was employed to smooth the overlapping part of the images.Finally,a primitive microscopic imaging system was built for testing the proposed methods.The results showed that this system yielded an imaging resolution higher than 2.19 um.In both cases of lightly out of focus and heavily out of focus,this system colud successfully focus.The testing results showed a focusing accuracy of 3.6 um(to meet the focal depth of 6.04)and a focusing success rate of 95%.After completing the autofocus tests,a full-field microscopy image stitching experiment was performed to compare the proposed methods with the existed methods in the literature.The results of both subjective and objective evaluations showed that the proposed methods are superior to the existed methods and that they could satisfy the requirements of high-resolution full-field microscopy imaging.