Design Patent Image Retrieval Based on Shape and Color Features

Based on the concrete characteristics and application background of design patent, according to the deficiency of the present retrieval system of design patent database, this paper proposed a synthesis design patent image retrieval method based on shape and color features. First, we present a shape retrieval approach based on the morphological description moment invariants, which can not only reflects the morphological characteristics of design patent image, but also has translation, rotation and scale invariability. Second, an improved color retrieval method based on the key local fuzzy color histogram which contains the colors’ statistic and local distributing characteristics is used to improve the color retrieval accuracy. At last, set a weight coefficient with the consideration of the similarity degree of the shape and color features. The experimental results indicated that the proposed method can retrieve similar design patent images fast and efficiently, and improve the recall ratio and precision of design patent image retrieval.