Segmentation of Sonar Images Based on Adaptive Thresholding with Image Histogram

With the advancement of technology, the imaging sonars have become the reality and their usage has been extensive in the area of obstacle avoidance in respect of Autonomous Underwater Vehicle (AUV). The underwater environment being heterogeneous, the sonar images have a very complex background, low contrast, and deteriorative edges. These characteristics, therefore, pose the difficulties for extracting the objects from the sonar images. In this paper we have reviewed the various existing image processing techniques in respect of the sonar images and discussed their shortcomings. After discussing the existing image processing techniques and their limitations, an algorithm has been proposed for processing these sonar images effectively and the results have also been compared with the results of existing techniques. Extracting the obstacle (objects) aspects such as range, bearing, size, shape, speed and course from the images received from Sonar are very important for the AUV in order to avoid the collision from the obstacles those may come into its path. Another very important criterion is the time taken to process the image, which must be as least as possible in order to provide more time for the AUV to take evasive action. For achieving this, proper segmentation of sonar images is a very important step in order to identify the objects (or obstacles) correctly in the least possible time. Several algorithms have been developed in the past for segmentation of images, however these methods did not provide the desired results when subjected to the real sonar images. Therefore a new segmentation method for processing the underwater Sonar images was developed by taking into account the available techniques and domain knowledge. This method is based on the thresholding of the image in which the threshold is calculated adaptively on iterative basis by taking the parameters of image histogram into consideration. The initial threshold value for each individual region of image has been selected by taking the histogram of the region under consideration. Adaptive thresholding utilizes a local window for each individual pixel and computes the new intensity value, based on the local histogram defined in the local window. This is then followed by the morphological, dilation and erosion operations before producing the final segmented image. The performance of the proposed algorithm has been compared with the other existing methods such as Edge detection, Adaptive Thresholding, Fuzzy C Means Clustering (FCM) and Adaptive Histogram Equalization. The results have also been presented in the tabulated form in addition to the segmented images. It has been concluded from the results that the proposed segmentation method achieves better segmentation results in respect of sonar images and is also highly efficient as it takes the least time for segmentation amongst all the methods.