An automated, feature-based framework for seabed mosaics

An automated, feature-based computer vision algorithm has been developed by a research team at the Institute for Microelectronics and Microsystems, Italy, which combines a sequence of calibrated underwater images into a 2D (two dimensional) seamless panorama for underwater mosaic construction. The algorithm comprises a statistical method that offers properly selected correspondences in the registration process, ensuring a low registration error for high-quality mosaics and a semi-real-time performance. The application of the contrast-limited adaptive histogram equalization (CLAHE) emphasizes the acquired scene structures, both by the increase of the dynamic range of the intensity levels image and reduction of nonuniform illumination and low-contrast effects. The use of a robust estimation technique for correspondences evaluation in CLAHE-enhanced images provides for efficient performance in the registration, avoiding the use of a global adjustment. The algorithm also helps obtain correct matches between adjacent images, even if low textured regions and a restricted superimposition are there.