Enhanced SURF- and Wavelet-Based Underwater Image Stitching

Underwater images are used in numerous scientific applications in the fields of marine geology, archaeology, military reconnaissance, finding underwater resources, detection of temporal changes under the sea, environment damage assessment, etc. Stitching these underwater images to obtain a clear view is a challenging and interesting problem for researchers. Several advances have been made to stitch normal images, but the problem of stitching underwater images has been poorly exploded because underwater images suffer from poor visibility conditions. Since underwater images are captured by unmanned underwater vehicles (UAVs), the orientation of the images obtained also introduces an difficult problem. An effective underwater image stitching technique is proposed in this paper. The images obtained from a particular location are oriented in correct angle using self-organizing map (SOM). The features of the oriented images are obtained with the help of speeded-up robust feature (SURF) registration technique. Hessian matrix plays the role of obtaining the feature points because it augments the number of feature points. From the obtained feature points, the overlapping regions in the images are identified. These regions are then eliminated using random sample consensus (RANSAC) algorithm. Finally, the pre-processed images are fused to obtain the overview of that particular area, thus provides a helping hand for researchers in various fields.