Abstract: In image database systems with content based retrieval, it is one of the key issues to extract features of images efficiently. The Image Semantic Data Model（ISDM） is reviewed, and a strategy for image analysis based on ISDM is presented in the paper. The main idea is to get heuristics from image classes definitions to guide features extraction. Using this method, the known techniques for image analysis are utilized whenever possible, and the extension of the feature extraction algorithm is supported to adapt various applications and improvements of computer vision techniques.