2010 Fiscal Year Final Research Report
Segmentation-free non-linear registration of image
| Project/Area Number |
20500159
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Perception information processing/Intelligent robotics
|
|---|
| Research Institution | The University of Aizu |
|---|
Principal Investigator |
OKA Ryuichi The University of Aizu, コンピュータ理工学部, 教授 (60347242)
|
|---|
| Project Period (FY) |
2008 – 2010
|
| Keywords | 画像理解 / マッチング / スポッティング / 動的計画法 / レジストレーション |
|---|
| Research Abstract |
The methods of image registration has been suffered from solving problems of parameter optimization of a model which is manually constructed. Moreover the segmentation is another difficult problem to be solved. These problems are commonly shared among the algorithms based on so-called model-based methods in image registration. In order to overcome these difficulties, we proposed algorithms based on appearance-based method which does not need any parameter to describe a model. A set of sub-images allocated in an image becomes a model to be registered on an input image. Our algorithm is based on so-called 2-dimensional dynamic programming which carries out full-pixel matching between images and also realizes so-called spotting, namely, segmentation-free optimal non-linear matching. Non-linear property includes two kinds of allowance of non-linear variation, namely, allocation of sub images and deformation of each sub-image. The experimental results showed that the proposed image registration works very well in the scheme of appearance-based method.
|
Research Products
(16 results)
