At present,the efficiency of change detection from high resolution remotely sensed imagery is generally not very high, and accuracy couldn't meet the practical requirements. A method which introduced frequency domain analysis is put forward in this paper. Firstly, it studies the distribution of image frequency energy. Using it as a basis, the image division is accomplished by summed-up distance matching function. Secondly, the definition and computation of texture regularity of change information(TRCI) clearly and accurately expresses the change information according to equal energy rule of image blocks, greatly narrowing the search area. Thus, the background information could be greatly suppressed so as to improve the efficiency and precision of detection. Finally, through measuring and comparing of the accuracy and efficiency of three different detection algorithms before and after introducing the preprocess, we find that it could provides more accurate results on the existed basis.