FPGA-accelerated Design And Implementation Of Optical Non-intuitive Super-resolution Microscopy Imaging Algorithms

Polarization parameter Indirect Microscopic Imaging(PIMI)is the frontier technology to break the diffraction limit and realize optical super-resolution microscopy imaging.It can be used for imaging of spatial change of material structure and image analysis of multidemension.PIMI algorithm needs to overlay multiple image data from the far field to reverse perform one image data from the near field.The software implementation method takes nearly 10 seconds to calculate one graphical result from the near field,which is much longer than the acquisition time of N images from the far field,so it cannot achieve real-time acquisition.PIMI algorithm varies with the physical properties of the light sources and optical elements in the actual optical path system.Usually,different models can only be completed in different systems,which is not convenient for practical application.In this paper,PIMI hardware platform is built based on SOPC system.The heterogeneous computing architecture composed of NIOS II and FPGA logic is used to build PIMI hardware and software co-data processing system.It not only realizes the multi-model hardware system platform,but also improves the computing speed by more than 50%.Specifically:1.As the data needs manual caching and cannot be run directly,this paper builds a hardware platform of PIMI algorithm.The parameter model of the experiment was input on the PC side,and the data was automatically cached in DDR2 off the FPGA chip,and PIMI algorithm was run directly in NIOS II;2.As the complex calculation of huge data,this paper adopts heterogeneous computing method.By converting PIMI algorithm,some part of the algorithm are suitable for FPGA,which are separated from NIOS II.Hardware has the characteristics of unique parallelism and pipelined processing,which can realize algorithm acceleration;3.For multi-application platforms,this paper divided the algorithm module according to the influence of model and the influence of experimental parameters,when building the module.The influence value of experimental parameters can be flexibly configured through NIOS II.The influence value of the model can only be changed when replacing the implementation model,so that the hardware architecture can be flexibly applied to multiple platforms.Based on this platform,PIMI algorithm can be applied to industrial detection,bioengineering,new medical treatment and other fields.