User­ Vs. Self-tuning Optimization: A Case Study on Image Registration

Abstract

In the last few years, there has been an increased interest in providing new soft computing (SC) algorithms, e.g., evolutionary algorithms (EAs), in which it is not needed the tuning of the control parameters. Usually, this new approach is named self-tuned and several models have been proposed to date. In our study, we aim to, firstly, describe how self-tuning EAs work and, secondly, provide a computational comparison of algorithms dealing with a specific computer vision (CV) task, named image registration (IR), as baseline. In particular, the development of automated IR methods is a well-known issue within the CS field and it was largely addressed from multiple viewpoints. IR has been applied to a high number of real-world scenarios ranging from remote sensing to medical imaging, artificial vision, and computer-aided design. Then, in this contribution it is provided a comprehensive analysis of the self-tuning approach by means of the experimental comparison of several EA-based IR algorithms proposed in the state-of-the-art.