The titles of ukiyo-e artworks often contain rich named entity information that closely corresponds to specific visual regions within the artwork. Effectively linking these entities across textual and visual modalities enables more accurate named entity recognition (NER) and facilitates deeper interpretation of historical materials. In this paper, we propose a multimodal NER method that integrates textual input, visual features from object detection, and external knowledge from large language models. Our method achieves state-of-the-art results in both NER for ukiyo-e titles and grounding of named entities to their corresponding image regions, thereby advancing the structured understanding and accessibility of ukiyo-e contents.