Abstract: Line segment feature is widely used in the field of image analysis and processing, and line segment extraction is an important step in pattern recognition and image matching which are feature-based. The pros and cons of the line segment extraction method directly affects the complexity and the processing effects of high-level image processing. However, there were some problems in most of the current line segments extraction methods. For instance, the endpoints of the extracted line segment were not accurate enough and the extracted line segments were always discontinuous. A HOG-feature-based approach was represented to overcome the above two problems. In order to extract line segments efficiently, in this method, the integration on rows and columns of a matrix was used to realize the projection of line segments with particular direction in the original image. And then with the application of the property of the rectangular function's first derivative, the line-segment endpoints positions were determined via the column vector's derivative. Theoretical analysis and the results of experiments verified that this method can not only extract straight segments accurately, but also solve the straight segment fracture and endpoints inaccurate problems existing in the classic methods to some extent.