For à trous wavelet-based remote sensing image fusion, it is significant to determine the number of decomposition levels (NDL) to derive the best tradeoff between the spatial and spectral qualities of the fused image. The tradeoff is modeled as a cost function from a novel three-dimensional variational point of view. By solving this cost function, a new fusion scheme is proposed that can be used to customize the tradeoff by weighting the wavelet planes with a factor. The optimal NDL is decided by the intersection of the characteristic curves of the spatial and spectral qualities. Three pairs of remote sensing images are used to test the analysis. The experimental results show that the proposed method provides the practitioner with a reference for selecting the fused image with different spatial and spectral qualities for different applications.