MULTI-FOCUS IMAGE FUSION: A COMPARATIVE STUDY

Abstract

Multi-focus Image fusion is a technique that has been studied a lot by researchers for years and every year a new multi-focus image fusion technique has been developed which surpasses the predecessor technique. Multi-focus image fusion methods combine two or more source images that have different focus points to generate a clear image. Multi-focus image fusion can be used in a lot of real-world applications such as remote sensing, medical imaging and surveillance, it is important to understand the fusion algorithms and to do a comparative research on them. In this paper is presented a review of different methods that have been developed during the years. In addition to multiple traditional multi-focus image fusion techniques this paper also presents recent advancements in multi-focus image fusion, including deep-learning approaches. I have evaluated different methods using two datasets that contain pairs of source images with different focus point and evaluated the performance of each algorithm by visually and objectively analyzing them. The objective evaluation of the fusion results is done by using a set of quality assessment metrics. The visual evaluation is done by comparing visually, the fused images of the method’s algorithms selected in this study. In my findings it is indicated that deep learning models for multi-focus image fusion considerably surpass the traditional techniques in terms of image quality and are trained for large image datasets.