Combining image and text features for medicinal plants image retrieval

This paper proposes a new approach for Indonesian medicinal plant image retrieval by combining leaf image and text features. Fuzzy Local Binary Patterns were used to extract texture features based on leaf image of medicinal plants. To improve the image similarity, we proposed Probabiistic Neural Network to calculate the weight of image features. The text features were extracted from medicinal plant documents in Indonesian language. Experiments result show that combining image and texture features in medicinal plant image retrieval improves the performance. The Average Precision (AVP) has increased from 0.3138 to 0.7081.