| Project/Area Number |
15300048
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | KYUSHU UNIVERCITY |
|---|
Principal Investigator |
NIIJIMA Koichi Kyushu University, Faculty of Information Science and Electrical Engineering, Professor, 大学院システム情報科学研究院, 教授 (30047881)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
OKADA Yoshihiro Kyushu University, Faculty of Information Science and Electrical Engineering, Associate Professor, 大学院システム情報科学研究院, 助教授 (70250488)
KUZUME Koichi Yuge National College of Maritime Technology, Information Science and Technology Department, Professor, 情報工学科, 教授 (80225151)
TAKANO Shigeru Kyushu University, Faculty of Information Science and Electrical Engineering, Research Associate, 大学院システム情報科学研究院, 助手 (70336064)
|
|---|
| Project Period (FY) |
2003 – 2006
|
| Project Status |
Completed (Fiscal Year 2006)
|
| Budget Amount *help |
¥15,800,000 (Direct Cost: ¥15,800,000)
Fiscal Year 2006: ¥3,300,000 (Direct Cost: ¥3,300,000)
Fiscal Year 2005: ¥4,100,000 (Direct Cost: ¥4,100,000)
Fiscal Year 2004: ¥3,200,000 (Direct Cost: ¥3,200,000)
Fiscal Year 2003: ¥5,200,000 (Direct Cost: ¥5,200,000)
|
|---|
| Keywords | biorthogonal wavelet / dyadic wavelet / lifting scheme / free parameters / learning / multi-resolution analysis / person identification / リフティングウェーブレット / 方向余弦最大化学習 / 分散最大化学習 / カートシス最小化学習 / 対象物検出 / 表情認識 / 顔画像認識 / リフティング双直交ウェーブレットフィルタ / 顔部位 / 逐次学習 / 追跡 / リフティングダイアデックウェーブレットフィルタ / 検出 / リフティングスキー / ノイズ除去 |
|---|
| Research Abstract |
Lifting wavelet is called the second generation wavelet, which is developed by Wim Sweldens of Lucent Technologies' Bell Labs. The lifting wavelet has a lifting term which incorporates controllable free parameters. Lifting scheme is a set of the down sampling filters constructed by adding the lifting term to initial biorthogonal wavelet filters. The constructed filters are also biorthogonal wavelet filters and keep the perfect reconstruction. These are very important properties for the lifting wavelet filters.
In this research, using such properties, we proposed a new learning method for determining the free parameters in the lifting term and developed a tracking system of moving objects based on the learned lifting scheme. And then, the various learning methods to determine the lifting parameters were produced by our lifting dyadic wavelet scheme which is extended version of the biorthogonal lifting scheme. Using the learned lifting scheme, we developed the image extraction algorithm and constructed person identification system via facial images captured by video frames. Furthermore, using the dyadic lifting scheme, we presented the technique for designing biorthogonal wavelets by dyadic wavelet. In addition, we studied to generate 3D objects by utilizing the wavelet multi-resolution analysis.
|
