Research on Thangka Image Retrieval Algorithm Based on Mean Hashing

Thangka images are rich in content and complex in texture information, so it is difficult to fully extract effective features and achieve good retrieval effect. This paper proposes that the extracted convolutional layer features are hashed with mean value encoding for Thangka image retrieval. First, VGG16 model is applied to extract the features of the last convolutional layer of the image for feature preservation, then the extracted features are mapped by mean hash algorithm, and finally, hamming distance is used for image similarity calculation. Experiments show that Thangka image retrieval based on mean hashing of image features is significantly better in terms of precision, recall and average retrieval accuracy than that based on image features of Thangka convolutional layer.