2022 Fiscal Year Final Research Report
Stochastic non-decomposition based tensor restoration and its application to image and signal processing
| Project/Area Number |
19K04377
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 21020:Communication and network engineering-related
|
|---|
| Research Institution | Chiba Institute of Technology |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2019-04-01 – 2023-03-31
|
| Keywords | テンソル復元 / 画像処理 / 信号処理 / 確率的最適化 / テンソル |
|---|
| Outline of Final Research Achievements |
Multidimensional signals captured by cameras and sensors, as well as Internet traffic data, may contain degradations such as missing data and noise. Such degradation is a major factor preventing effective use of signals and data, i.e., recognition and knowledge acquisition from the data. To solve this problem, we proposed a method to significantly improve the computational complexity and memory usage of the tensor restoration algorithm without sacrificing the restoration performance by applying a stochastic optimization framework to the non-decomposition tensor restoration algorithm. This framework achieves low-rank tensor approximation with low computational complexity. As a result, the proposed method outperforms existing methods in image restoration and Internet traffic restoration.
|
| Free Research Field |
信号処理
|
| Academic Significance and Societal Importance of the Research Achievements |
本課題によって得られた研究成果の学術的意義としては非分解型・確率的テンソル復元を実際の問題に応用した多数のアルゴリズムを提案したことや，その際に用いたテンソルノルムの非凸拡張間の関係を明確にしたことなどが挙げられる．また，このような学術的成果は今後，非分解型テンソル復元を自動運転などに用いられる画像認識AIのための前処理や，医療，天文学などの幅広い分野にも応用可能であることから，社会的意義も大きいといえる．
|
