Study On Fast Autofocusing Method In Coherent Diffraction Imaging System

Autofocusing technology is an essential automatic adjustment technology that relies on the clarity evaluation function to achieve the clear and accurate imaging in imaging system.This technology originated from traditional optical microscopes and has been widely promoted in the new generation of microscopic imaging systems,as well as in the more extensive application of computational microscopy imaging technology.Microscopic imaging is closely related to autofocusing technology,from the focusing process of the microscope stage to the calculation of distance parameters in computational imaging technology.Achieving fast and intelligent autofocusing performance in microscopic imaging systems is a pressing research task for the disciplines that apply computer vision automatic detection systems.The development history of autofocusing technology is analyzed,and the classification,characteristics and implementation of three current technologies in the field of autofocus,namely region of interest extraction,sharpness evaluation function and intelligent search algorithm are summarized in this thesis.Starting from the precise determination of the image distance of the optical microscopic imaging system and the calculation of the diffraction distance of the microscopic imaging system,we aim to solve the practical problems such as the contradiction between the focusing accuracy and the fast calculation in the automatic focusing process of the microscopic imaging system,and the poor robustness.In response to the above research tasks,the following research work is carried out in this thesis:(1)To meet the demand for fast and accurate autofocusing in the small depth-offield range of optical microscopy for high-resolution imaging,an autofocusing method for color microscopy imaging in optical microscopy is studied.To address the problems of inaccurate and low sensitivity of autofocusing results due to complex color changes in color microscopy imaging,an autofocusing method based on the Tanimoto coefficient is proposed.The pixel values of the out-of-focus dataset two-color channel images after normalization and thresholding are used as the evaluation objects of pixel differences,and the Olympus microscope and the simple autofocusing microscopy experimental platform are used to acquire the out-of-focus image stacks as the experimental data,respectively.The experimental results fully demonstrate the advantages of the new method in terms of noise robustness,focal point recognition sensitivity and algorithm efficiency.Based on the correlation between pixel difference and distance under single wavelength illumination,it is verified that the new method is still valid for microscopic imaging systems with monochromatic illumination.(2)A multi-distance coherent diffraction imaging technique is studied for the extraction of regions of interest in multiple diffraction planes for fast and accurate autofocusing.To address the drawback of high computational load of multi diffraction plane autofocusing,a significant feature region extraction algorithm based on the region of maximum density of corner points is proposed.In order to reduce the computational load,the algorithm is used to extract the most representative feature regions in the sample as the input of clarity evaluation function.In the sharpness evaluation algorithm,the multi-directional maximum gradient is used as the image sharpness criterion to improve the autofocus sensitivity by making full use of the multi-directional features.The design of unequal interval recording of diffraction intensity images and iterative strategy are used to achieve fast,sensitive,and accurate autofocusing and sample reconstruction for multi-distance coherent diffraction imaging system.(3)For multi-distance coherent diffraction imaging,the large computational effort of auto-focusing mainly comes from the uniform search(traversal search)strategy.To address this problem,an intelligent search strategy based on subdivision search is proposed to achieve fast and accurate acquisition of diffraction distances of multiple diffraction images.The intelligent search algorithm can provide a more accurate and efficient search method for the autofocus technique,which in turn improves the imaging quality and efficiency of the microscopic imaging system.The new method is 30 times faster than the traversal search with the same search range and search accuracy.(4)The problem of simultaneous determination of object and image plane positions in a single-lens coherent diffraction imaging system is solved.The existing clarity evaluation function cannot calculate the objective and image plane positions in the imaging system simultaneously.The mean square error and structural similarity are used as the clarity evaluation scheme.The two cases of known and unknown image plane positions are discussed,and the feasibility of the new method is confirmed by the fact that the additional scattering sheet in the imaging system facilitates the improvement of the autofocusing performance.Meanwhile,a new cascaded iterative engine model is designed for sample image reconstruction.The research work in this thesis provides solutions for automatic focusing techniques and sample reconstruction for optical microscopy and computational microscopy imaging systems,laying the foundation for the development of lightweight,portable and highperformance optical microscopy imaging systems.