Joint Visual and Semantic Information Image Retrieval

In image retrieval, there exists a "semantic gap" problem, which refers to the discrepancy between the image semantics understood by computers through extracted image features and the image semantics perceived by humans. This paper studies an image retrieval algorithm based on image visual information and semantic information. The algorithm utilizes image features combined with hash encoding to describe the visual characteristics of images, and employs a word vector generation model to convert semantic words of images into word vectors, thereby describing the semantic features of the images. To better combine the similarities of visual features and semantic features, the algorithm establishes a reranking retrieval matrix for image searching. The paper validates the effectiveness of the proposed algorithm through experiments, demonstrating that the retrieved images highlighted by this algorithm are consistent in terms of semantic information while ensuring the similarity of visual features.