A Feature Extraction Method Invariant to General Coordinate Transformation

• Project/Area Number: 17K00245
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Perceptual information processing
• Research Institution: Osaka City University
• Principal Investigator: Toriu Takashi 大阪市立大学, 大学教育研究センター, 特任教授 (80347484)
• Project Period (FY): 2017-04-01 – 2020-03-31
• Project Status: Completed (Fiscal Year 2019)
• Budget Amount: ¥2,600,000 (Direct Cost: ¥2,000,000、Indirect Cost: ¥600,000); Fiscal Year 2019: ¥520,000 (Direct Cost: ¥400,000、Indirect Cost: ¥120,000); Fiscal Year 2018: ¥520,000 (Direct Cost: ¥400,000、Indirect Cost: ¥120,000); Fiscal Year 2017: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
• Keywords: 一般座標変換 / 計量テンソル / 不変性 / スムージング / 画像処理 / 変換不変性 / 特徴抽出 / コンピュータビジョン / アフィン変換 / 微分演算子 / アフィン不変な評価関数 / 画像 / 情報基礎 / 人工知能 / 幾何学 / アルゴリズム

Outline of Final Research Achievements

Images are deformed for various reasons such as movement of the viewpoint and distortion of the lens. It is desirable that the image processing result is invariant to such deformation. In general, image deformation is represented by general coordinate transformation. In this research, I developed a smoothing method that is invariant to general coordinate transformation and demonstrated its effectiveness.
Specifically, we introduced a metric tensor that is transformed appropriately to the general coordinate transformation and defined an evaluation function that is invariant to the general coordinate transformation using the metric tensor. Furthermore, we derive a differential equation to minimize the evaluation function and find that it is invariant to general coordinate transformation. The image is smoothed by its differential equation.

Academic Significance and Societal Importance of the Research Achievements

アフィン変換に不変な画像処理手法は数多く提案されている。しかし、一般座標変換に対して不変なスムージング手法の開発は他に例をみない。この手法はスムージングだけでなく、一般座標変換に対して不変な特徴抽出など、他の画像処理にも展開できることが期待される。
画像は、視点の移動、レンズの歪みなどのさまざまな理由で変形を受ける。提案手法によって、画像が受ける様々な変形に不変な画像処理が実現し、画像処理のロバスト性が向上する。それによって、画像処理の応用場面の拡大が期待される。

Research Products

• [Journal Article] A Method for Extracting Affine Invariant Local Features Using a Metric Tensor (2017)
  • Author(s): TORIU TAKASHI
  • Journal Title: Fifth International Conference on Advances in Computing, Communication and Information Technology - CCIT 2017 Volume: 1 Pages: 48-54
  • DOI: 10.15224/978-1-63248-131-3-18
  • Peer Reviewed / Open Access
• [Journal Article] An Adaptable Low Dimensional Image Generation Model for Eliminating Illumination Influence (2017)
  • Author(s): TAKASHI TORIU、HIROSHI KAMADA 、THI THI ZIN
  • Journal Title: Sixfth International Conference on Advances in Computing, Communication and Information Technology - CCIT 2017 Volume: 1 Pages: 40-46
  • DOI: 10.15224/978-1-63248-138-2-09
  • Peer Reviewed / Open Access
• [Presentation] General Coordinate Invariant Image Smoothing Using a Metric Tensor (2020)
  • Author(s): Takashi Toriu
  • Organizer: 2020 3rd International Conference on Image and Graphics Processing
  • Int'l Joint Research
• [Presentation] Image smoothing which is invariant under general coordinate transformation (2018)
  • Author(s): Toriu Takashi
  • Organizer: Proceedings Volume 11069, Tenth International Conference on Graphics and Image Processing (ICGIP 2018)
  • Int'l Joint Research
• [Patent(Industrial Property Rights)] 図形の特徴抽出装置 (2017)
  • Inventor(s): 鳥生 隆
  • Industrial Property Rights Holder: 大阪市立大学
  • Industrial Property Rights Type: 特許
  • Filing Date: 2017