Abstract: Image Database (IDB) is a combination of the traditional database and the image processing techniques. An intelligent retrieval method of image database, which is guided by artificial intelligence, is presented. Pattern recognition and image processing techniques, heuristic knowledge are all taken into consideration in this method. This method uses an image feature vector to represent an image and builds a retrieval tree of the image base through clustering to seek the optimal image with best-first search for users on it. How to obtain a heuristic evaluation function by using expert's experience and sample training is focused on. These conditions for the existence of evaluation function is presented as well.