Building Accelerated Denoising Algorithms for Graph Signal Processing

2020 Fiscal Year Final Research Report

• Project/Area Number: 18K18076
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 61010:Perceptual information processing-related
• Research Institution: Waseda University
• Principal Investigator: Sugimoto Kenjiro 早稲田大学, 理工学術院(情報生産システム研究科・センター), 講師(任期付) (00773483)
• Project Period (FY): 2018-04-01 – 2021-03-31
• Keywords: 画像処理 / 高速化 / バイラテラルフィルタ / ノンローカルミーンズフィルタ / ハイパースペクトル画像 / デノイジング

Outline of Final Research Achievements

We have developed efficient denoising algorithms for various data structures (high-dimensional images, hyperspectral images, point clouds, graph signals, etc.) that are widely used in engineering. In particular, we analyzed the compatibility between computational complexity and accuracy in terms of both theoretical and practical performance, and refined various filter methods.
In this research period, we have published 1 international journal article, 1 international letter, 22 conference presentations (including 11 international conferences and 2 invited lectures), and received 2 awards. Most of the international conference presentations were accepted by the most prestigious flagship conferences in signal/image processing field (ICIP and ICASSP), and also included best papers and awards for young researchers. From these highly-evaluated outcomes, our research showed high impacts and potentials in signal/image processing communities.

Free Research Field

画像処理

Academic Significance and Societal Importance of the Research Achievements

画像撮像デバイスの急速な発展に伴い，撮像データはますます複雑化している．これらに対するデノイジングなどの画像処理手法は通常多大な計算量を必要とし，それが高度な手法を導入する際の障壁となっている．我々はこういった高度な手法を現実的な時間で処理できる方法を，理論と実用（ソフトウェアやハードウェア）の両面から分析し，より広い対象への応用の実現に取り組んでいる．本テーマでの成果は，こういった多様なデータ構造に対する画像処理アルゴリズムの効率化の基礎を担う重要な処理であり，またその一部は現状で世界最高性能を達成している．