AUTOMATIC AND CONCURRENT DETERMINATION OF OPTIMAL VALUES OF NONLOCAL MEANS FILTERING PARAMETERS BASED ON BAYESIAN FORMULATION IN IVUS IMAGES

摘要

A challenging issue that has emerged regarding nonlocal means filtering based on Bayesian formulation in order to remove noises from ultrasound images is the determination of the optimal parameters of this filtration without the involvement of users concurrently before the start of the filtration. This issue is effective in both the filtration quality and the time required for process. This paper, presents an algorithm based on optimization algorithm in order for automatic and concurrent determination of the optimized parameters of this filter. The intravascular ultrasound (IVUS) images used for experiments and evaluation of the results have been taken from seven different individuals. The resolution of these images is 500 * 500 pixels that is prepared by a transducer at a frequency of 30 Mhz at 0.55 m/s. The proposed algorithm works by combining nonlocal means filtering based on Bayesian formulation with genetic optimization algorithm. The cost function of the genetic algorithm is introduced based on the intended filter. In order to evaluate the algorithm proposed in this paper, a two-stage evaluation process was. At the first stage examinations, 50 IVUS images were recorded in 50 experiments of the applied proposed algorithm. The values of evaluation criteria and output images were then recorded and organized in a descending order. Ten preliminary groups of these parameters that had the best results in terms of the intended criteria were chosen for the second stage experiments. The second stage includes 30 experiments on 30 images not seen so far by the proposed algorithm. The details of the results are provided within the paper text extensively together with the obtained tables. In this paper, we have tried to simplify the way nonlocal means filtering should be used to improve the quality of intravascular ultrasound (IVUS) images. To this end, the algorithm proposed in this research calculates the optimal values of the parameters of this filter before filtration automatically with no user involvement. It also carries out the filtration using the optimal values of parameters that are the outputs of the proposed algorithm thereby increasing the quality of filtration by optimized parameters.