ECDL 2006, Alicante, September 18, 2006 Naga Srinivas Vemuri, Ricardo da S. Torres, Rao Shen, Marcos Andre Goncalves, Weiguo Fan, and Edward A. Fox A Content-Based.

ECDL 2006, Alicante, September 18, 2006

Naga Srinivas Vemuri, Ricardo da S. Torres, Rao Shen, Marcos Andre Goncalves, Weiguo Fan, and Edward A. Fox

A Content-Based Image Retrieval Service for Archaeology Collections

query image

searching results

Comments/Questions?

Runnable notes

Log onto ETANA @ http://feathers.dlib.vt.edu:8080/etana/servlet/Start Go to http://feathers.dlib.vt.edu:8080/cbisc_etana/ Pick a sample image from the folder

http://feathers.dlib.vt.edu/~etana/sampleim Actual input is a ppm image, just for viewing purposes, I included jpg

image too. Select a ppm image, select BIC descriptor, and k=20, hit go The resultant page returns a set of DL records based on image content. Explain [View complete record], [Add to Items of Interest] and [Share

Item] Click only [Share Item] and explain the page..Note: You can’t click back button on browser page, so I suggest clicking

[share item] as it has the most information.

Scenario Dr. Doug Clark is an archaeology professor at Sierra University. He

received a pottery image through email from an archaeological excavation site Bab edh-Dhra. He visits the ETANA website, and tries the content-based image search on the image archive of the site present in ETANA-DL. The CBIR service returns a brief description of digital objects whose image content is similar to the query image supplied.

He browses through the set of digital library records returned to him, and identifies an image of a DL record closely resembles the pottery image retrieved. He looks at the brief description of the DL record, and using the [Add to Items of Interest] option, adds it to the list of records of interest. Then, using [Share Item] option, he shares this particular record with the archaeological field expert who sent him this email. Using [View Complete Record] option, he discovers the excavation site details of the returned record. Based on these details, he suggests the field expert, possible locations to excavate for more interesting discoveries.

Future Work Another DL service that we plan to enrich using

this content based retrieval service is the DL Recommendation component. Collaborative filtering based recommendation algorithms tend to perform badly in the presence of sparse user interests data. In such situations, we boost the recommendation component with the recommendations generated by the content based retrieval service. Also, we plan to explore other image descriptor algorithms for better results.