Multi-focus image fusing based on non-negative matrix factorization

Abstract

Multi-focus image fusion is a process of obtaining a new all in-focus merged image from two or more partially defocused images of the same scene and same imaging condition. The merged image includes the information of the original images and improves the reliability and intelligibility for object detection and target recognition. The most widespread methods for image fusion are wavelet transform based methods. However, the facts that the original pixel values of input images are not preserved in the fused image and different multi-scale image fusion schemes will lead to different results cause that the wavelet methods present a limited quality performance compared with a cut and pasted fusion reference model. In this paper, a new multi-focus image fusion approach is proposed based on non-negative matrix factorization (NAIF). The cut and pasted fusion scheme is adopted in the new fusion approach. Cut the source images into small-size blocks, factorize the corresponding image blocks using NAIF, pick out the sharpest blocks according the NAIF coefficient, and combine them as an in-focus image. The experiment results show that the proposed approach outperforms the wavelet based fusion methods, both in visual effect and objective evaluation criteria.