Comparative Study on Enhance Side Scan Sonar Images to Improve Object Detection Efficiency

Side Scan Sonar (SSS) is an important tool in oceanography, providing high-resolution images of the sea floor and underwater structures. Oceanographic studies, which include the study of underwater topography, object detection, and mapping of marine habitats, in conjunction with the use of machine learning or deep learning, still face problems in image processing because the input of the original image data is not complete or clear enough to help produce highly accurate processed results. These imperfections can make underwater object detection challenging. To address this issue, researchers have conducted a comparative study to enhance SSS images and improve object detection efficiency. This research proposes five strategies for improving images from side scan sonar by leveraging advanced image processing techniques. This helps maintain color balance affected by signal strength variations, such as histogram equalization, which addresses image color and brightness balance, and multiscale retinex with color restoration, which manages image brightness. And use statistical tools to analyze the signal, such as Rayleigh Distribution, and enhance and improve image quality by adjusting pixel intensities to filter out noise. Block-matching with 3D filtering uses non-local similarity and collaborative filtering to reduce noise caused by low light; it also includes adjusting brightness appropriately with perceptual linearization. Enhancement: The evaluation from the YOLOv8 model will indicate the performance of each algorithm based on the objective of detecting artificial corals. The results show that MSRCR has the highest performance in enhancing the image with a mAP of 76%.