After any tsunami, hurricane, flood, or any other natural disasters, immediate assessment and response are important for the minimization of damage and protection of human lives. Underwater image processing has emerged as one such crucial tool in disaster recovery and emergency response situations. It is used for the assessment of damage caused to underwater and coastal infrastructure, tracking environmental changes, and supporting rescue operations. However, underwater imaging is very often adverse conditions of low visibility, turbidity, and low light, which extremely hampers the acquisition of sharp, actionable images during such events. This chapter introduces an Underwater Image Enhancement Approach Using Recurrent-based Convolutional Neural Networks (RCNN) to enhance the image clarity and supply accurate real-time data for disaster recovery in smart cities. The RCNN model improves underwater image quality by eliminating effects such as blurring, noise, and contrast decay, which are often seen in similar images. This will help make assessments concerning the degree of damage sustained by various underwater structures like bridges, dams, and pipelines more effective. The improvement provided will also help the rescuers and recovery teams in real-time imagery of submerged areas in order to navigate safely during rescue missions and identify survivors or obstacles. Hence, the enhanced underwater images provide good information for post-disaster planning by gaining further insight into modifications in terrain below water, displacement of sediments, and erosion of the coastline. With smart city infrastructure allied with advanced image processing techniques, the approach provides an effective response to a disaster and the planning period for long-term resiliency so that greater restoration and safety are achieved within a shorter timeframe after the occurrence of a natural disaster.