Performance evaluation of adaptive neuro fuzzy system (ANFIS) over fuzzy inference system (FIS) with optimization algorithm in de-noising of images from salt and pepper noise

Abstract

In all image processing task image pre-processing is very important. It includes image enhancement and image de-noising. It is the process of removing unwanted information from an image. Image de-noising plays very important role in all image processing based research. An adaptive neural network (ANN) plays a vital role in many research areas such as image processing, character recognition, forecasting and hand writing recognition etc. An ANN is very much useful in machine learning and problem solving. Now a day, neuro fuzzy system is famous in research because it is easy enough in handling the nature of complex problems and imprecise data. The main objective of the paper is removing impulse noise from the image using adaptive neuro fuzzy inference system (ANFIS). It combines the learning abilities of neural network with inference capability of fuzzy system. Membership function plays vital role in fuzzy system. Different types of membership function are there, among which for our research we have used the triangular membership function. The proposed network is used to find different noisy pixel patterns. The fuzzy techniques with the network filter the noise without affecting the textures and fine details of the image. From the experimental results it is concluded that ANFIS perform better than fuzzy inference system (FIS) with optimization algorithm. The PSNR value is used as a performance metric. The proposed method is evaluated on standard test images. The result of ANFIS system is compared with other noise restoration methods such as mean filter, median filter, fuzzy system and fuzzy system with optimization algorithm. The experimental result shows that our proposed method achieves the maximum PSNR of 62.510 db which is high when compared with other presented methods.