A clustering based approach to efficient image retrieval
read more
Citations
Exploiting Deep Features for Remote Sensing Image Retrieval: A Systematic Investigation
A Pattern Similarity Scheme for Medical Image Retrieval
Efficient fingerprint search based on database clustering
Automatic image annotation by an iterative approach: incorporating keyword correlations and region matching
Traffic Sign Classification Using Ring Partitioned Method
References
Ten lectures on wavelets
Content-based image retrieval at the end of the early years
Query by image and video content: the QBIC system
Related Papers (5)
Frequently Asked Questions (12)
Q2. What is the natural choice in color indexing?
The natural choice, according to the typical soft computing literature [20], is to impose a smooth decay of the resemblance function when the inter-color distance increases.
Q3. Why does FUZZYCLUB use a secondary clustering technique?
In order to improve the query processing time, FUZZYCLUB incorporates a secondary clustering technique to “pre-organize” the database to significantly save the search time.
Q4. What is the p-value for the texture vector?
L2 distance metric is applied to the texture vector and the shape vector, respectively:qp pq T TTd −= (9)qp pq S SSd −= (10)where pT , pS are the texture feature vector and the shape feature vector for region p, respectively, and qT , qS are for region q, respectively.
Q5. What is the k-means algorithm used to cluster the feature vectors?
Consequently the k-means algorithm [17] is used to cluster the feature vectors into several classes with every class in the feature space corresponding to one spatial region in the image space.
Q6. What are the common CBIR systems?
Typical region based CBIR systems include Berkeley Blobworld[15], UCSB Netra[16], Columbia VisualSEEK[9], and Stanford IRM[6], of which [9,15,16] are the classic region based CBIR systems which require significant user interaction in defining or selecting region features, preventing from a friendly interface to users, especially to non-professional users.
Q7. How long does it take to index the images?
The whole indexing time, including running the secondary clustering after indexing each image, for the 2000 image database takes 60-70 minutes, corresponding to about 2 seconds per image.
Q8. How many images are returned to each of the three systems?
For each group images in the 2000 database images, the authors randomly select 30 images as queries to each of the three systems, respectively.
Q9. What is the k-means algorithm for image indexing?
Within each region, the authors define three types of features: color, texture, and shape, along with the conventionalgeometric information as the feature vector for image indexing.
Q10. Why are the recall values not proportional to the precision values in this case?
Since the number of relevant images in the database for each query image is the same, the recall values are not computed as they are proportional to the precision values in this case.
Q11. Why is it important to capture the inaccuracy in color indexing?
On the other hand, due to its inherent nature of “inaccuracy” in description of the same semantic content by different the color quantization and/or by the uncertainty of human perception, it is important to capture this “inaccuracy” when define the features.
Q12. What is the definition of a content-based image retrieval?
Content-based image retrieval (CBIR) concerns automatic or semi-automatic retrieval of image data from an imagery database based on semantic similarity between the imagery content.