Abstract: The concept of granular computing into line detection in remote sensing images using Hough transform is introduced, according to the characteristic of granular computing that it can observe problems at multiple views and levels. An algorithm on line detection based on granular computing is established by analyzing linear edge features with granularity of various sizes. Different quantification interval of corresponding granular level is employed, so as to extract lines with different length and width efficiently. An application on airport runways extraction in the real IKONOS image proves the validity of this algorithm.