An effective shape-based retrieval algorithm for design patent images

作者：Cao Jiangzhong^*; Dai Qingyun; Cao Lu; Li Xuming

来源：Journal of Computational Information Systems, 2010, 6(3): 773-781.

• Bounding box
• CBIR
• Contour shape
• Data sets
• Database images
• Descriptors
• Fourier descriptors
• Furniture design
• Image content
• Object region
• Other algorithms
• Patent database
• Region shape
• Shape descriptors
• Shape features
• Shape-based retrievals
• Similar design
• Zernike moments

It's very significant to retrieve design patents according to the image content. An effective shape-based retrieval algorithm for design patent images is proposed in this paper. The proposed algorithm first performs a preprocessing to obtain the normalized object region. Then, a new shape descriptor based on the bounding box-contour distances (BBCD features) is used to represent the contour shape and the improved grid feature is used to describe region shape. Finally, the similarity between the query and database images is computed according to the two types of shape features. Users can retrieve similar design patent images according to the sorted similarity. The proposed algorithm has been used to retrieve design patents from the furniture design patent database which contains about 500,000 patent images and has been compared with two other algorithms-Fourier Descriptor and Zernike Moments Descriptor, which are widely used to represent shape. The experiment results show that our proposed algorithm is effective and outperforms the two other algorithms on the furniture design patent dataset.

• 出版日期2010
• 单位中山大学