2003 Fiscal Year Final Research Report Summary
Separation of diffuse and specular components of surface reflection by using Independent Component Analysis
| Project/Area Number |
14580439
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | National Institute of Advanced Industrial Science and Technology (AIST) |
|---|
Principal Investigator |
UMEYAMA Shinji National Institute of Advanced Industrial Science and Technology (AIST), Neuroscience Institute, group leader, 脳神経情報研究部門・脳情報工学グループ, 研究グループ長 (90344100)
|
|---|
| Project Period (FY) |
2002 – 2003
|
| Keywords | Independent Component Analysis / Mutual Information / Diffuse Reflection / Specular Reflection / Dichromatic Reflection Model / Polarizer |
|---|
| Research Abstract |
(1)We made an experimental observation system with a high-precision digital video camera and a rotating polarizer which is controlled by a computer. By using this system, we made a separation experiment of diffuse and specular reflection components of an object. The separation result was good, and we presented this result in MIRU2002. In this experiment we developed a new separation algorithm which uses a special property of the task, such as the problem includes only two components. The proposed algorithm computes the mutual information between estimated two components directly. Since it is a kind of enumeration algorithm, it is stable and we can set algorithm parameters very easily.
(2)We made a simultaneous observation system of two images in order to make our method applicable to cases that objects or lights move. In this system we used a half-mirror prism. The light from the object is split into two rays by the prism, and they are observed by two different video cameras equipped with polarizers of different polarizing directions. This observation system simultaneously gives two images with different amount of specular and diffuse components. This system can capture 16 pairs of images per second. This result was presented in SSIIO9.
(3)The proposed separation method appeared in IEEE Trans. on PAMI in April 2004.
|
