Title: A Medical Image Fusion Method based on MDLatLRRv2

Abstract: Since MDLatLRR only considers detailed parts (salient features) of input images extracted by latent low-rank representation (LatLRR), it doesn't use base parts (principal features) extracted by LatLRR effectively. Therefore, we proposed an improved multi-level decomposition method called MDLatLRRv2 which effectively analyzes and utilizes all the image features obtained by LatLRR. Then we apply MDLatLRRv2 to medical image fusion. The base parts are fused by average strategy and the detail parts are fused by nuclear-norm operation. The comparison with the existing methods demonstrates that the proposed method can achieve state-of-the-art fusion performance in objective and subjective assessment.