Research On The Blind Restoration Methods Of Microscopic Image

In recent years, microscopic image restoration researches are playing an increasinglyimportant role in biomedical, neural science, cell and molecular biology and other fields.They become one of the effective ways to solute the image quality degradation problem.Because blind image algorithms need less prior knowledge of image, they become a hardand hot spot of people’s research. This dissertation focus on the study of microscopicimage restoration algorithms, mainly from two points of views to study, which aremaximum a posteriori probability and the total variation image restoration algorithm. Themain works is listed as follows:(1)The traditional Maximum a posteriori method is studied, and on this basis, focusedon the Generalized Gaussian distribution model. Considering the Generalized Gaussiandistribution model has the nature of simulating all kinds of noises that encountered inengineering model. Combine both of them. The front part of the model with Poissondistribution instead of, and the rear part with generalized gauss model instead of.Simulation results show that, compared to the traditional RL (Richardson-Lucy) algorithm,this method restored image detail more clearly, the overall visual effect is better.(2) Combination traditional method of maximum a posteriori probability and the totalvariation regularization term. Considering the gradient of image, then the large gradientarea is the edge region, it needs to maintain the image detail; else, the small gradient area isthe texture region, it needs to reduce the ringing and suppress the increase of noise.According to the gradient of different regions this section uses different regularization term.The division of the gradient using the thinking of the golden section point. Experimentsshow that running time by this improved algorithm is half than that by the traditional RLalgorithm, and the peak-noise ratio is improved, and other indicators have improved.(3)The traditional total variation blind image restoration has been studied, joinedWeber Law and regularization term; this section improves the total variation image blindrestoration algorithm. It can be seen by experiments that the image quality improved bythis algorithm is better than that of the original algorithm, and the image is clearer.This dissertation simulates the three algorithms using the methods of completingsimulated pictures and the actual microscopic images. Simulation results show the valueand effectiveness of those algorithms.