Cell Image Fusion Methods Based On Adaptive Focusing Process

Cytological detection,as a current major disease detection method,occupies an important position in the medical field.The premise of cytological research is to have cell samples,and the collection of cell samples cannot be separated from the microscope.In the early days when image processing technology and sensor technology were not yet developed,image acquisition was completely completed by humans,and the objective of focusing was achieved by constantly adjusting the objective and eyepieces manually.This method is time-consuming and labor-intensive.As the working time increases,the physical strength and concentration of the physician gradually decrease,and the accuracy of the focus will be further affected.On the other hand,because it is an artificial subjective judgment whether the focus position has been reached,there is a certain gap between the judgment standards and basis of each physician,which will also cause a certain discrepancy between the focus results.Aiming at this kind of defect,this thesis has studied the cell image fusion method based on adaptive focusing process.Firstly,the platform is built,and then an adaptive autofocus algorithm based on the gray value of background pixels is proposed for experimental analysis.This method is constructed by two new definition evaluation functions combined with mountain climbing method.Although it can focus quickly and accurately,and has high sensitivity and good timeliness,it can not solve the problem of moving the microscope back and forth in the focusing process.In view of this disadvantage,this paper further proposes a microscope focus prediction method based on neural network.According to the phenomenon that the gradient value of pixels changes almost the same after the microscope moves for a certain distance,a definition evaluation function based on the gradient value change is proposed,which can be directly used to establish the prediction model instead of being normalized.The definition evaluation function value collected in the later stage is directly input intothe prediction model for focus coordinate prediction.Then a large number of experiments are carried out to verify the performance of this method.In the process of focusing image acquisition,because of lens imaging and objective field of vision,there will be multi-focus phenomenon.In order to get a clearer image,this paper proposes a region fusion method based on block segmentation.In this method,block based fusion and region based fusion are combined to reduce the false segmentation phenomenon in the whole image segmentation process,reduce the difficulty of region registration,and eliminate the "block effect".Finally,the regions that need to be fused in the block are fused together to get a clear image.Finally,The cell image fusion method based on adaptive focusing process is verified by simulation experiments and compared with some classic methods to analyze the feasibility and superiority of the method.Experiments show that the method proposed in this thesis can quickly,accurately and clearly capture images.