Abstract: Linear feature in remote sense image has great geonomy and geometry significance, linear feature ascertaining and extracting in high resolution remote sense image is the basis and key premise of high-level image project such as image analyzing and understanding. In this paper, we discussed a method based on direction and frequency selected through designing frequency domain filtering for the remote sense image linear feature automatic detection and recognition, first owed to Fourier transformation method transform the image into frequency domain, and analyzed the relations between linearity characteristic and its spectral line, also the relations between the linearity characteristic and image frequency, at the bases of the direction and frequency selected above the Gabor wave filter for image linearity feature extracting was constructed. The relevant extract experiments was tested with Quickbird image, the experiments result indicates that the linearity characteristic being extracted in this method is fairly good, it provided a reference for high-resolution remotely sensed imagery linear feature extracting, and also provided a new ways for feature extract in frequency domain use of high-resolution satellite imagery based on accurately frequency spectrum analysis.