Breast cancer diagnosis using digital mammogram is a practical field of investigation. The positive results could
affect the mortality ratio of human life. In this paper, a new model is proposed by using a curvelet transform as a per process for feature extraction and classification of mammogram images. Curvelet based texture features yielded high accuracy rate to classify mammogram images. This was expected since the curvelet transform is able to capture the multi-dimensional features in wedges. The proposed system covers two problems: the first is abnormality detection. And the Second is to distinguish the abnormal tissues into between benign and malignant. The classification accuracy rate achieved to distinguish between normal and abnormal is 97.03%, while it is 91.68% to distinguish between benign and malignant. These results indicate that using curvelet based texture feature can improve the classification of mammogram.