Development of an acceleration framework for image processing using architectures with tensor processing

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K24368
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Multi-year Fund
• Review Section: 1002:Human informatics, applied informatics and related fields
• Research Institution: Tokyo University of Science
• Principal Investigator: Maeda Yoshihiro 東京理科大学, 工学部電気工学科, 助教 (80843375)
• Project Period (FY): 2019-08-30 – 2022-03-31
• Keywords: 画像処理 / 行列演算機構 / GPU / FIRフィルタリング

Outline of Final Research Achievements

We investigated how to effectively utilize matrix arithmetic instructions, which can realize fast matrix sum-of-products operations, in image processing algorithms. In this study, we studied image filtering algorithms, which is a fundamental process in image processing, suitable for matrix arithmetic instructions. The results of this study showed that the matrix operation instructions could be accelerated image filtering algorithms.

Free Research Field

画像処理

Academic Significance and Societal Importance of the Research Achievements

本研究は，行列の積和演算を高速に計算可能な行列演算機構を画像処理にも応用する方法を明らかにするものである．行列演算機構は，深層学習に特化した専用演算器であったが，本研究により，様々な処理においても行列演算機構が有効に活用できることを示した．これにより，行列演算機構の多種多様な用途での活用が期待される．