Sonar image segmentation and quality assessment using prominent image processing techniques

Abstract

Sound navigation and ranging (SONAR) is a technique that uses sound propagation to navigate, communicate or to detect objects on or under the surface of the water. It even forecasts what lies beneath the sea waves. It uses sound waves to observe the happenings occurring on the region of seafloor. Sonar images that are framed using sound waves are not directly suitable for analysis as it would be affected primarily by a speckle noise which misguides the way of interpretation. The metaphors of substances in those images are very important in the field of oceanography to suit many applications. The scope of this research work is to promote the quality of an image in order to get a better interpretation than using a raw image. This article also deals with the novel work to identifying the type of noise present in sonar image using an artificial neural network (ANN) and the overall accuracy of the network achieved is found to be 93.37%. In addition, filtration of noise has been done using different digital filters and it is proved that the combination of SRAD and Gaussian filter is best suited for sonar image denoising processes. Finally, this work contributes many ideas related to the segmentation of objects in an image using various techniques like Fuzzy C Means (FCM), K means, Expectation–maximization (EM) algorithm, Otsu thresholding and Toboggan Segmentation. The research outcomes are compared with different quality matrices and it exhibits that the FCM segmentation has high accuracy than the other methods and all the results were validated through MATLAB software.