Vegetation fractional cover which defines the amount of vegetation on the surface of the land, is a key parameter in land surface models. Based on a geometric-optical model in combination with linear spectral mixture analysis, this paper discusses the estimation of shrub fractional cover in Wushen Banner of Inner Mongolia in the Mu Us Sandland using HJ-1B image. Firstly, we acquire the surface reflectance after geometric correction and atmospheric correction of the HJ image. Then we assume that the reflectance of the mixed pixel is a simple linear combination of two components, and further calculate the areal proportion of the illuminated background within each pixel based on linear spectral mixture analysis. After that, combined with the measured shrub structural parameters, the shrub fractional cover is estimated using an inverted geometric-optical model. Finally, the result is validated through the measured shrub cover of 13 sample plots. The R2 of the result is 0.898 with the root mean-squared error of 0.136 which indicates the reliability of the combined method.