DPF - a perceptual distance function for image retrieval
read more
Citations
A survey of content-based image retrieval with high-level semantics
Manifold-ranking based image retrieval
Efficient manifold ranking for image retrieval
Generalized Manifold-Ranking-Based Image Retrieval
Human activity recognition for video surveillance
References
Support vector machine active learning for image retrieval
MindReader: Querying Databases Through Multiple Examples
Similarity, interactive activation, and mapping
Comparing discriminating transformations and SVM for learning during multimedia retrieval
Maximizing expected generalization for learning complex query concepts
Related Papers (5)
Frequently Asked Questions (14)
Q2. How many transformations are performed for each image in the 60; 000-image set?
For each image in the 60; 000-image set, the authors perform 24 transformations including scaling, downsamping, cropping, rotation, and format transformation.
Q3. What are some of the common distance functions used to measure similarity between image vectors?
Various distance functions, such as the Minkowski metric, earth mover distance [5], and fuzzy logic, have been used to measure similarity between feature vectors representing images.
Q4. What is the purpose of this project?
In this project, the authors mine visual data extensively to reverse-engineer a good perceptual distance function for measuring image similarity.
Q5. What is the metric used to measure image similarity?
A variant of the Minkowski function, the weighted Minkowski distance function, has also been applied to measure image similarity.
Q6. What is the purpose of this paper?
Through empirical study, the authors demonstrate that DPF is very effective in finding images that have been transformed by rotation, scaling, downsampling, and cropping, as well as images that are perceptually similar to the query image (e.g., images belonging to the same video shot).
Q7. What is the effect of m on the distances between two objects?
When m < p, it counts only the smallest m feature distances between two objects, and the influence of the (p m) largest feature distances is eliminated.
Q8. How is the weighted Minkowski distance function defined?
By assigning each feature a weighting coefficientwi (i = 1 p), the weighted Minkowski distance function is defined asdw(X;Y ) = (pXi=1wijxi yij r)1 r : (2)By applying a static weighting vector for measuring similarity, the weighted Minkowski distance function assumes that similar images resemble the query image(s) in the same features.
Q9. How does DPF perform in the visual data mining?
The authors discovered the dynamic partial distance function (DPF) through mining a large set of visual data, and showed that DPF outperformed the traditional functions by significant margins.
Q10. What is the first stage of the project?
In the second stage, the authors freeze the features to discover a perceptual distance function that can better cluster similar images in the feature space.
Q11. What is the main topic of this article?
Research in content-based image retrieval has steadily gained momentum in recent years as a result of the dramatic increase in the volume of digital images.
Q12. What is the goal of the project?
In other words, their goal is to find a function that can keep similar images close together in the feature space, and at the same time, keep dissimilar images away.
Q13. What is the distance between the two objects?
when r is set as 2, it is the well known Euclidean distance; when r is 1, it is the Manhattan distance (or L1 distance).
Q14. What is the effect of m on the dissimilar images?
In Section 3, the authors will show that DPF makes similar images aggregate more compactly and locate closer to the query images, simultaneously keeping the dissimilar images away from the query images.