To improve the reliability of reference map preparation for scene matching, it is necessary to analyze the matching performance of remote sensing image. Remote sensing image matching performance metric was proposed based on independent component analysis. Firstly, texture basis functions are produced based on independent component analysis and a set of probability functions that describe the coefficients distribution corresponding to each of the texture basis functions are calculated to extract matching regions. Secondly, the area ratio index, distribution index and stability index for matching regions are defined. Lastly, remote sensing image matching performance metric is constructed based on the three indexes. The experiment shows that the proposed remote sensing image matching performance metric index is highly correlated to real matching probability. It improves the reliability of reference map preparation and can meet the need of remote sensing images selection for scene matching.