A New Model for Image Fusion Using Bi-dimensional Empirical Mode Decomposition with Directional Information

Abstract

Image fusion is a procedure in which two or more images of one scene captured by different sensors are combined into one image. The target of image fusion is to produce images which are more suitable for human visual perception and further machine-processing. Empirical Mode Decomposition (EMD) is a new tool for signal (especially non-stationary signal) analysis. Comparing to other multi-scale transforms, EMD is more suitable for detecting information of images. Therefore, we propose a new model for image fusion based on Bi-dimension EMD (BEMD). In addition, considering that the existing BEMD can hardly capture the directional information which is important for image fusion, directional filter bank will be employed to run on each level coefficients in the step of image decomposition. The proposed method is tested on medical images and compared with some newly-proposed image fusion models. The experimental results show that the proposed method outperforms others in terms of several fusion evaluation metrics.