A new development with Gaussian processes in probabilistic model based image processing
| Project/Area Number |
26330046
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Statistical science
|
|---|
| Research Institution | Hiroshima City University |
|---|
Principal Investigator |
SUEMATSU Nobuo 広島市立大学, 情報科学研究科, 准教授 (70264942)
|
|---|
| Project Period (FY) |
2014-04-01 – 2017-03-31
|
| Project Status |
Completed (Fiscal Year 2016)
|
| Budget Amount *help |
¥4,290,000 (Direct Cost: ¥3,300,000、Indirect Cost: ¥990,000)
Fiscal Year 2016: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2015: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2014: ¥2,470,000 (Direct Cost: ¥1,900,000、Indirect Cost: ¥570,000)
|
|---|
| Keywords | 画像分割 / ガウス過程 / 混合モデル / 空間相関 / 画像処理 / 正規混合モデル / EMアルゴリズム |
|---|
| Outline of Final Research Achievements |
Image segmentation is a process to divde a given image into segments corresponding to objects or some meaningful areas. We have developed a new image segmentation algorithm which uses Gaussian processes (GPs), and it accomplished significant improvements in segmentation accuracy. GPs are used to model spatial correlations in images. Although existing methods uses Markov Random Fields (MRFs) for the purpose, GPs have some advantages over MRFs. For example, GPs can specify spatial correlations more directly and more flexibly than MRFs can. Moreover, GPs can model spatial correlations independently of
image resolution.
|
Report
(4 results)
Research Products
(8 results)
