Research on distortion-tolerant, controllable, parametric image matching
| Project/Area Number |
26330207
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Perceptual information processing
|
|---|
| Research Institution | Hosei University |
|---|
Principal Investigator |
WAKAHARA Toru 法政大学, 情報科学部, 教授 (40339510)
|
|---|
| Project Period (FY) |
2014-04-01 – 2017-03-31
|
| Project Status |
Completed (Fiscal Year 2016)
|
| Budget Amount *help |
¥2,340,000 (Direct Cost: ¥1,800,000、Indirect Cost: ¥540,000)
Fiscal Year 2016: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2015: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2014: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
|
|---|
| Keywords | パターン認識 / 変形耐性画像マッチング / 画像マッチング / 変形耐性 |
|---|
| Outline of Final Research Achievements |
1. A new technique of 2D projection transformation (PT) invariant template matching, GPT (Global Projection Transformation) correlation, was proposed. The GPT correlation method determines optimal eight parameters of PT that maximize a normalized cross-correlation value between an input image and the PT-superimposed template. 2. Recognition experiments made on the well-known MNIST handwritten digit database via a combination of the GPT correlation and k-NN techniques achieved the lowest error rate of 0.30% ever reported for k-NN based classification. 3. The programs in C language with source codes used in the above-mentioned experiments were published on the Web. 4. The GPT correlation method was greatly enhanced to stabilize and accelerate convergence to the optimal solution of eight parameters via strict formalization of the objective function and refinement of its computational model.
|
Report
(4 results)
Research Products
(6 results)
