ISSN : 2319-7323





INTERNATIONAL JOURNAL OF COMPUTER SCIENCE ENGINEERING


Open Access

ABSTRACT

Title : MULTI-ADAPTIVE PARALLEL NEURO FUZZY INFERENCE SYSTEM FOR MEDICAL IMAGE DENOISING
Authors : Dr.D.Revathi
Keywords : Fuzzy rules, Image Denoising, Membership Function, Neural Network,
Issue Date : Jan-Feb 2020
Abstract : Normally, images are corrupted by impulse noise during acquisition and transmission over communication medium. As a result, noise removal and enhancement are essential in digital image processing. Since, the performances of subsequent processes in image processing are strictly dependent on the success of the image denoising. However, the complexity of this process is high due to the noise removal operator with the requirement of preserving useful information in the image during noise removal process. In previous researches, OFPTFLIS (Optimized Firefly based Parallel Type-2 Fuzzy Logic Filtering System) is introduced for removing the noise from the images. In this approach, image denoising is performed in optimized parallelized manner by using the firefly optimization and type-2 fuzzy logic system. However, the processing speed of the denoising process is less when multiple fuzzy rules are considered. Hence in this research, Multiple Adaptive Parallel Neuro Fuzzy Inference System (MAPNFIS) is proposed. This approach is introduced for improving the processing speed of denoising process with the multiple parameters. In addition, the fuzzy membership functions selection process is achieved adaptively. Here, Adaptive Fuzzy Inference System (AFIS) is hybridized with the Neural Networks (NN). It is a combination of both back propagation and the least square algorithms. When multiple membership functions are used, the performance of denoising is speeded up by computing the weighting factors of the rules. This system makes the utilization of a hybrid-learning rule for optimizing the fuzzy system parameters of a first order Sugeno system. Finally, the experimental results show that the proposed approach achieves better performance than the other approaches in terms of PSNR, MSE, MAE and computation time.
Page(s) : 80-86
ISSN : 2319-7323
Source : Vol. 9, No. 1