Several classic image fusion methods were analyzed by using the extended model of general component substitution (GCOS) framework. The relations between the injection modes of these fusion methods were investigated. On this basis, a novel spatial variable detail injection method, which took the correlations of panchromatic and multispectral bands into consideration, was established. By combining with different spatial detail extraction methods, three novel high-pass modulation (HPM) image fusion schemes were proposed. The fusion results of Ikonos and Quickbird data were analyzed quantitatively by using the proposed methods, and the result was compared with those by the traditional HPM. Fusion results of visible bands show that the fusion performance is improved by the proposed modification of modulation coefficients and detail extraction methods compared to HPM, and the contribution of the improved modulation coefficients is more significant. However, the fusion results of near infrared (NIR) bands indicate that the study of the filter design should be paid more attentions for the fusion in NIR and panchromatic bands.