The robustness of registration measures is one of the most frequently considered problems in image registration process. The robustness of some dominant registration measures to overlay area changing is discussed by using statistic method. In addition,quantitative analysis is included. The results indicate that the linear correlativity between these registration measures and overlay area is very prominent. Among these registration measures,the normalized mutual information [H(A)+H(B)]/H(A,B) has the smallest variation coefficient while Absolute Difference(AD)has the largest coefficient. Image registration using this normalized mutual information measures can reduce the performing time and speed up the image registration process in the condition of ensuring accuracy,which has been validated in the registration tests.