This paper proposes a novel multi-spectrum to RGB method with direct structure-tensor reconstruction. The goal of the proposed method is to generate a fused color image which preserves structure-tensor of the original multispectral image while keeping naturalistic color of a reference RGB image. Existing multi-spectrum to RGB methods generate the fused image by sequential pipeline that consists of fused-gradient estimation and reintegration. The existing sequential pipeline tends to generate the fused image whose structure-tensor is deviated from the structure-tensor of the original multispectral image. Contrary to the existing sequential pipeline, the proposed method generates the fused image directly with the structure-tensor of the original multispectral image by minimizing an energy which consists of the structure-tensor term and the reference color constraint term. The proposed direct structure-tensor reconstruction enables us to preserve the structure-tensor of the multispectral image while keeping the naturalistic color of the reference RGB image. The experimental results on multispectral images including visible and NIR images show that the proposed method outperforms the existing methods in terms of not only image quality but also quantitative evaluations.