With the popularity of multimedia applications, the huge amount of image and video related to real life have led to the proliferation of emerging storage techniques. Contented-based image retrieval and classification have become attractive issues in the last few years. Most researches concerning image classification focus primarily on low-level image features (e.g. color, texture, shape, etc.) and ignore the conceptual associations among the objects in the images. In this paper, we propose a new image classification method by using multiple-level association rules based on the image objects. The approach we proposed can be decomposed of three phases: (1) building of conceptual object hierarchy, (2) discovery of classification rules, and (3) classification and prediction of images. At the first phase, we use a hierarchical clustering method to build the conceptual hierarchy based on the low-level features of image objects. At the second phase, we devise a multi-level mining algorithm for finding the image classification rules. The classification task is performed at the last phase. Empirical evaluations show that our approach performs better than other approaches in terms of classification accuracy.