Loading...
Please wait, while we are loading the content...
Similar Documents
Perceptual Consistency Improves Image Retrieval Performance (2001)
| Content Provider | CiteSeerX |
|---|---|
| Author | Leow, Wee Kheng Long, Huizhong |
| Abstract | An ideal retrieval system should retrieve images that satisfy the user's need. Therefore, it should compare the similarity between images in a manner consistent with human's perception. However, existing computational similarity measurements are not necessarily consistent with human's perception. This paper proposes an approach of improving retrieval performance by improving the perceptual consistency of computational features and similarity measurements. A novel method for incrementally measuring perceptual distances and constructing perceptual space based on relevance feedback is presented. Only a small number of relevant judgments is required in each feedback. Feedbacks from multiple queries are accumulated and incrementally updates the measurements of perceptual distances between images. As more and more feedbacks are received, the constructed space becomes more consistent with human's perception. Test results show that the distances measured by this incremental method correlates very well with perceptual distances measured in psychological experiments. Moreover, the more perceptually consistent are the distance measurements, the better is the image retrieval performance. |
| File Format | |
| Publisher Date | 2001-01-01 |
| Publisher Institution | In Proc. ACM SIGIR, 2001 |
| Access Restriction | Open |
| Subject Keyword | Similarity Measurement Computational Feature Distance Measurement Psychological Experiment Relevant Judgment Perceptual Consistency Computational Similarity Measurement Multiple Query Small Number Constructed Space Novel Method Image Retrieval Performance Incremental Method Correlate Retrieval Performance Perceptual Distance Relevance Feedback Ideal Retrieval System Perceptual Space |
| Content Type | Text |
| Resource Type | Article |
