Multibeam Imaging Sonar Image Fusion via -total Variation Model

Abstract

Multibeam imaging sonar provides detailed information that has proven crucial in various applications like cognitive robots, navigation, and inspection. However, because of hardware limitations of multibeam imaging sonar, the captured image with controlled frequency suffers from the dispersion and attenuation with respect to specified material. To integrate more information, a more accurate description about the same scene can be obtained via a technique known as image fusion. Currently, total variation regularization method provides an efficient way to achieve this task on gradient domain. In this paper, we propose a new problem formulation with l₁-total variation model for multibeam sonar image fusion. An alternative minimization method is proposed, which can integrate dual-frequency imaging sonar image into an enhanced one. To the authors’ best knowledge, this is the first study concerning multibeam sonar image fusion. Experimental results on a real world dataset illustrate the effectiveness and efficiency of the proposed algorithm.