This paper proposes a content-based manga retrieval system that addresses three key challenges in manga search: image description, retrieval-localization, and query modality. The system uses a manga-specific image-describing framework, including efficient margin labeling, edge orientation histogram (EOH) feature description, and approximate nearest-neighbor search using product quantization (PQ). It also introduces a sketch-based interface for intuitive interaction with manga content, enabling sketch-based querying, relevance feedback, and query retouch. To evaluate the system, a novel dataset called Manga109 was created, consisting of 109 manga titles with a total of 21,142 pages drawn by professional manga artists. This dataset is the largest publicly available manga image dataset. The system was tested through a comparative study, localization evaluation, and large-scale qualitative study. The results showed that the proposed method outperforms existing methods in retrieval accuracy, can effectively localize object instances with reasonable runtime and accuracy, and that sketch-based querying is particularly useful for manga search. The system's image description process involves detecting candidate areas of objects using selective search and extracting EOH features. These features are then compressed into binary codes using PQ for efficient retrieval. The system also includes a sketch-based interface that allows users to interact with manga content through relevance feedback and query retouch. The interface enables users to refine search results by drawing additional sketches or modifying existing queries. The proposed system has been evaluated on the Manga109 dataset, which contains a wide range of manga genres and is publicly available for academic research. The results of the experiments confirmed that the system effectively addresses the challenges of manga retrieval, providing a more intuitive and efficient search experience for users. The system's ability to handle large-scale data and its effectiveness in retrieving relevant manga images make it a valuable tool for manga research and applications.This paper proposes a content-based manga retrieval system that addresses three key challenges in manga search: image description, retrieval-localization, and query modality. The system uses a manga-specific image-describing framework, including efficient margin labeling, edge orientation histogram (EOH) feature description, and approximate nearest-neighbor search using product quantization (PQ). It also introduces a sketch-based interface for intuitive interaction with manga content, enabling sketch-based querying, relevance feedback, and query retouch. To evaluate the system, a novel dataset called Manga109 was created, consisting of 109 manga titles with a total of 21,142 pages drawn by professional manga artists. This dataset is the largest publicly available manga image dataset. The system was tested through a comparative study, localization evaluation, and large-scale qualitative study. The results showed that the proposed method outperforms existing methods in retrieval accuracy, can effectively localize object instances with reasonable runtime and accuracy, and that sketch-based querying is particularly useful for manga search. The system's image description process involves detecting candidate areas of objects using selective search and extracting EOH features. These features are then compressed into binary codes using PQ for efficient retrieval. The system also includes a sketch-based interface that allows users to interact with manga content through relevance feedback and query retouch. The interface enables users to refine search results by drawing additional sketches or modifying existing queries. The proposed system has been evaluated on the Manga109 dataset, which contains a wide range of manga genres and is publicly available for academic research. The results of the experiments confirmed that the system effectively addresses the challenges of manga retrieval, providing a more intuitive and efficient search experience for users. The system's ability to handle large-scale data and its effectiveness in retrieving relevant manga images make it a valuable tool for manga research and applications.