Application of Perfect-code to hashing and mathematical analysis

2021 Fiscal Year Final Research Report

• Project/Area Number: 19H04140
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Review Section: Basic Section 61010:Perceptual information processing-related
• Research Institution: Kumamoto University
• Principal Investigator: Koutaki Gou 熊本大学, 大学院先端科学研究部(工), 准教授 (20582935)
• Co-Investigator(Kenkyū-buntansha): 白井 啓一郎 信州大学, 学術研究院工学系, 准教授 (00447723) ; 京地 清介 工学院大学, 情報学部(情報工学部), 准教授 (70634616) ; 松井 勇佑 東京大学, 大学院情報理工学系研究科, 講師 (80780676)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: バイナリハッシング / アダマール行列 / Perfect-Code / クロスモーダル

Outline of Final Research Achievements

Binary hashing has attracted attention as a means of efficient data compression/retrieval of multimedia data (images, text and video), which is increasing with the spread of SNS. Until now, the applicants have devised a hashing method using Hadamard code sequences and have shown its effectiveness. In contrast, in this study, hashing using Perfect-Code, which is more separable, and extensions to complex and quaternionic numbers were carried out. In fact, evaluation experiments using Wiki and MNIST have confirmed that the proposed method can improve the mAP significantly compared to the conventional HC-SDH.

Free Research Field

画像処理

Academic Significance and Societal Importance of the Research Achievements

これまでPerfect-codeをバイナリハッシングに利用する試みはこれまでなく学術的な新規性がある。さらにそれが情報検索精度の向上に役立つことを示すことができた。また、これらの手法が複素数・4元数へ拡張できること、および性能改善につながることを示せたことは意義が高いと考える。