In image retrieval, corner detection with auto-adaptive threshold and target area extraction based on curvature are proposed in this paper. The algorithm judges the importance of corners by using curvature. Firstly, it selects the true corners through the auto-adaptive threshold. Meanwhile, they enhance their curvature. Then, it determines the image's center of gravity by the corners with larger curvature. Last, it extracts the target area by regarding the center of gravity as the centroid. The image retrieval experimental results show that this algorithm can not only detect the corners and extract the target area effectively, but also improve the accuracy and efficiency of image retrieval compared with the traditional method. It provides a new approach for the retrieval of image with complicated background.