Research On Non-reference Evaluation Method Of Slide Scanning Image Sharpness

The applications of digital pathology in information management,remote consultation,expert consultation,slides interpretation communication and practical teaching all depend on clear and undamaged slide scanning images.The slide scanning imaging system technology is a key for the development of digital pathology,but cannot guarantee that the slide image is completely clear and non-destructive in the process of digitizing the slide,which will affect the subsequent interpretation and research.Therefore,it is necessary to evaluate the sharpness of the slide scanning image and find out the blurred image block after digitized.Among the evaluation methods,the subjective evaluation method is time-consuming and laborious.The reference evaluation method is limited because it is difficult to obtain the reference image.The current non-reference image quality evaluation method evaluates the multiple distortion characteristics of the image,and needs to use the image information in the image database.And the non-reference sharpness evaluation method can gives a reference value of sharpness,but for images of different content,the reference value fluctuates greatly.Therefore,the non-reference sharpness evaluation method is needed to evaluate the sharpness of the slide scanning image,which only relies on the information of the slide scanning image itself and is not easily affected by the content and the objective lens resolution.This thesis adopts the method of establishing "pseudo" reference images,and generates "pseudo" reference images through Gaussian blur,and combines the idea of image quality evaluation of structural similarity.This thesis designs and implements a non-reference sharpness evaluation method.This method sets the size of the Gaussian blur kernel according to the imaging resolution.After the image is sorted by blocks,the number of structural difference comparison areas is determined according to the detection result of the content volume and the imaging resolution.Then This thesis calculates the structural differences before and after Gaussian blur in these regions,and uses the average of the structural differences as the result of sharpness evaluation.This thesis designs and implements a non-reference image sharpness evaluation application program,which can browse,analyze and export the results of the evaluation of the sharpness.This thesis compares the performance of the method in this thesis and other methods.Because considering the imaging resolution and image content,the robustness of the method in this thesis is better than other general methods on the evaluation of the clarity of the slide image.On the basis of the subjective evaluation of the sharpness of the slide image,This thesis compares the subjective consistency of the method in this thesis and other methods.The experimental results show that the subjective consistency of the method in this thesis is better than other methods in evaluating the sharpness of the slide scanning image.