There is incompatibility between spatial characteristic enhancement and spectral information preservation in remote sensing image fusion. We propose an image fusion algorithm to reduce this incompatibility. It is based on Intensity-Hue-Saturation (IHS) and regularization in wavelet domain. For the high-resolution intensity image, the proposed approach assumes a wavelet domain local Gaussian model as prior distribution of the spectral characteristic, an Symmetric Conditional Markov(SCM) model as prior distribution of the spatial correlation, whose parameters are learnt from the analysis of the corresponding Pan wavelet coefficients. The constrained optimization problem is solved with the gradient descent algorithm. Visual and statistical results obtained by using Landsat-7 Enhanced Thematic Mapper Plus (ETM+) images demonstrate that the proposed method can improve the spatial characteristic while preserving the spectral information effectively.