| Project/Area Number |
11680385
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | SHIZUOKA UNIVERSITY |
|---|
Principal Investigator |
ABE Keiichi Shizuoka University, Professor, 情報学部, 教授 (80022193)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
SUGIYAMA Takahiro Shizuoka University, Assistant, 情報学部, 助手 (70293595)
|
|---|
| Project Period (FY) |
1999 – 2000
|
| Project Status |
Completed (Fiscal Year 2000)
|
| Budget Amount *help |
¥3,400,000 (Direct Cost: ¥3,400,000)
Fiscal Year 2000: ¥1,200,000 (Direct Cost: ¥1,200,000)
Fiscal Year 1999: ¥2,200,000 (Direct Cost: ¥2,200,000)
|
|---|
| Keywords | color image processing / color image segmentation / color coordinating system / color space / quantitative evaluation / region segmentation / region boundary / color perception / 色知覚特性 / クラスタリング手法 / ISODATA法 / カラーインデックス / エッジベクトル抽出器 / 色分類心理実験 |
|---|
| Research Abstract |
The main results of this research are as follows :
(1)Utilization of image space information in segmenting color images
We investigated how to use image space information in addition to color space information nad proposed a method which produces satisfactory segmentation results from human viewpoint with a set of fixed parameter values, independent of the input image.
(2)Comparative evaluation of color spaces for color image segmentation
We made a comparative evaluation of eleven color spaces, from the viewpoint of which color spaces are adequate for region segmentation of color images. For this purpose we made human experiments on color pixel classification and region segmentation, and utilize their results. We clarified which color spaces are adequate for using in color image segmentation and which are not.
(3)Quantitative evaluation of region segmentation results of color images
We proposed a method of evaluating quantitatively segmentation results of color images. The criteria considered are uniformity of colors in a region and contrast between two adjacent regions. We designed an overall evaluating function and compared derived function values with human evaluated scores. Except undersegmented results, function values satisfactorily close to human evaluation are obtained.
(4)Application of the above findings to color image applications
We applied the results and findings described above to various applications of color image processing.
|
