This paper presents a content-based manga retrieval system that addresses the challenges of manga image description, retrieval-localization, and querying. The system uses an objectness-based edge orientation histogram (EOH) feature description method and product quantization (PQ) for efficient nearest-neighbor search. A sketch-based interface is proposed for natural interaction, enabling sketch querying, relevance feedback, and query retouch. The authors also introduce the Manga109 dataset, consisting of 109 manga titles with 21,142 pages, which is the largest publicly available manga dataset. Evaluations using this dataset show that the proposed method outperforms previous methods in terms of retrieval accuracy, object localization, and sketch querying. The system is designed to be efficient, with a retrieval time of 70 ms on a single computer for the Manga109 dataset. The paper concludes by discussing future directions, including combining sketch and keyword-based searches.This paper presents a content-based manga retrieval system that addresses the challenges of manga image description, retrieval-localization, and querying. The system uses an objectness-based edge orientation histogram (EOH) feature description method and product quantization (PQ) for efficient nearest-neighbor search. A sketch-based interface is proposed for natural interaction, enabling sketch querying, relevance feedback, and query retouch. The authors also introduce the Manga109 dataset, consisting of 109 manga titles with 21,142 pages, which is the largest publicly available manga dataset. Evaluations using this dataset show that the proposed method outperforms previous methods in terms of retrieval accuracy, object localization, and sketch querying. The system is designed to be efficient, with a retrieval time of 70 ms on a single computer for the Manga109 dataset. The paper concludes by discussing future directions, including combining sketch and keyword-based searches.