Ultrasonic Image Denoise Analysis Based On Markov Random Fields

Due to the influences of various noises, the signal-to-noise ratio (SNR) and reliability of digital signal are low in ultrasonic non-destructive testing, and it is one of difficult tasks in NDT fields. In this thesis, the model and statistical properties of the structure noise are studied. And some methods of enhancing signals and suppressing noise are trying to be developed, in order to improve the SNR of the ultrasonic signal and qualities of the ultrasonic imaging. These methods are based on the Markov random fields.First, the theory of Markov random field is introduced. Selecting a proper neighborhood system and using the ability of Markov random field to describe spatial dependence, MRF can be used to model the structural and characteristic of images, analyze ultrasonic testing image. Selecting an appropriate model and making use of the optimal algorithm — simulated annealing to estimate the parameters, wonderful image denoising and restoration can be achieved. Then a Gauss-MRF denoising algorithm based on Markov random field model is presented and the refresh formula of relaxation labeling is deduced. According to Bayesian theory, the recovery of image denoise is transformed a problem of maximum a posteriori (MAP) estimation. Finally, the experiment conclusion is obtained based on MAP rule.Markov random field models, Gibbs distribution function and simulated annealing used together for ultrasonic images denoise processing are discussed internally in works. We can lean from the simulation and experiment that this method can effectively stand out the defect signals and restrain the noise, enhance SNR of ultrasonic testing images.