Evaluation

1
Off-line
On-line
Query Neural Image
Labeled Brain Images
1
Image Preprocessing (normalization,registration)
comparable image
Comparable images
2
Index Feature Extraction
Similarity
2
Feature Extraction
metric
Query image index features (5-6)
Each image has 50 potential index features
Image Retrieval (nearest neighbors)
4
Feature Selection (classification)
3
Labeled images
Unlabeled images
3
Classification
5-6 image features
Most discriminating feature subset Image index
feature vector Weighted Euclidean Similarity

5
Evaluation
2
Semantic Based Image Retrieval Result Evaluation
Retrieval Precision