Underwater image communication has been a key focus for the researchers in last decade. This is due to its potential applications like: military strategy, ecological preservation, scientific discovery and submerged cultural heritage protection. The real-time analysis of these images has the potential to facilitate and ameliorate search and rescue operations. Despite this, the underwater image communication suffers due to limited underwater acoustic channel bandwidth. In view of this setback, instead of transmitting the entire pixel-level image, segmented image transmission can be used to reduce the bandwidth requirements and communication time. To achieve the same, in this paper the side scan sonar (SSS) image segmentation is formulated as a clustering problem. The SSS images captured over time from an underwater vehicle are segmented using dynamic multi-objective optimization algorithm. The targeted algorithm used here is Dynamic Non-dominated Sorting Genetic Algorithm-II (DNSGA-II). This algorithm makes use of learning from previous population and prediction using support vector regression (SVR) for reduced computation complexity along with better convergence. The potency of this work is validated using a subset of images from benchmark Seabed objects KLSG-II dataset.