Research Article
Partially Occluded Facial Image Retrieval Based on a Similarity Measurement
Table 2
Performance comparison of the proposed method for partially occluded facial image retrieval in the case of occlusion by sunglasses.
| | Algorithm | Image format used in SIFT matching | Number of subregions | Retrieval rate |
| | Gabor-LBP [26] | Without SQI | 6 × 6 | 75.56% | | 10 × 10 | 75.56% | | With SQI | 6 × 6 | 75.56% | | 10 × 10 | 75.56% |
| | SIFT total weight [35] | Without SQI | 6 × 6 | 68.89% | | 10 × 10 | 45.93% | | With SQI | 6 × 6 | 77.04% | | 10 × 10 | 71.11% |
| | SIFT probe weight [36] | Without SQI | 6 × 6 | 59.26% | | 10 × 10 | 45.93% | | With SQI | 6 × 6 | 64.44% | | 10 × 10 | 53.33% |
| | Proposed method | Without SQI | 6 × 6 | 77.78% | | 10 × 10 | 92.60% | | With SQI | 6 × 6 | 94.07% | | 10 × 10 | 92.59% |
|
|