In recent years, due to the rapid development of 3D modeling technology, 3D model databases have been increasingly available on the Internet. More and more 3D models can be easily downloaded through the Internet. This has directly led to the development of 3D shape retrieval technology, in which the system needs to return a similar 3D model according to the user requirement. This paper presents a new 3D shape retrieval method, which takes a 3D model as input and the system automatically returns some models that are most similar to the input shape from the model database. For a given input model and every model in the database, first, generated a magnitude of 2D sketch images of the model from different perspectives by the computer. Next, foreach 2D sketch image generated, the algorithm applies Gabor filter to extract the local features of the image, and quantifies the features in order to obtain a histogram representing the sketch image. For each 3D model, we then obtain a number of histograms representing the model. Thus, by comparing the histograms of every two models, we can compute the similarity value between the two models, and so retrieve the most similar shape to the input shape. In brief, the proposed method is capable of extracting effective features of 3D model through 2D image analysis method and evaluating the similarity between models. Experimental results show that the proposed algorithm performs well on some public datasets.