We propose a new method for extracting urban rivers from high resolution remotely sensed imagery in the frequency domain. Radial and angular samplings are used to analyze the magnitude spectrum （frequency spectrum） signature of the urban river after discrete Fourier transformation of the remotely sensed imagery into the frequency spectrum. The bilinear features of the urban river in high resolution remotely sensed imagery are divided into two parts: edge features and low frequency information. The frequency recognition marks of these two parts are retrieved based on the periodic texture frequency spectrum model and the spectrum energy discipline of ground objects. Meanwhile

arc ring two dimensional log Butterworth band-pass filter and two dimensional Butterworth low-pass filter are designed to extract edge features and low frequency information

respectively. The two extraction results can be used in extracting urban rivers effectively from highresolution remotely sensed imagery. Evaluation of the extraction results shows that the proposed algorithm is effective in extracting urban rivers from high spatial resolution remotely sensed imagery.