Abstract: Texture character is an important clue of image analysis. There are two categories of texture analysis techniques-statistical methods and structural methods, both of which have their own advantages. Based on the two kinds of methods, a novel method is proposed, which describes linear image texture through a kind of ARG （attributed relational graphs）, and applies this technique into image retrieval. Because the image’s characteristics are presented intuitively through the structure of graph, this method has excellent texture description ability. Furthermore, straight-line segment is extracted as texture primitive. Three geometrical position attributes, which are invariant to changes of scale, rotation and translation between texture primitives, are put forward. Therefore the image recognition and retrieval applications of this method possess excellent anti-noise ability. Practical experiments show that the performance of the new approach is satisfying.