160 likes | 275 Vues
This paper presents a novel approach to Content-Based Image Retrieval (CBIR) by integrating spanning trees and normalized cuts. We address the limitations of traditional methods in fast segmentation and user interaction. The proposed system allows users to input a query image, quickly segments it, and retrieves similar images through progressive refinement. The hybrid segmentation technique combines hierarchical image segmentation with rapid grouping, resulting in improved speed and quality. Our method shows promising results for both image and video retrieval applications.
E N D
Combining Spanning Trees and Normalized Cuts for Internet Retrieval Sharat Chandran1 and Abhishek Ranjan2 ViGIL IIT Bombay www.cse.iitb.ac.in/graphics/ 1. 2.
Need for CBIR • 1994 : Yahoo Text based search • 2002 : Google image search search for “Apple” • A related problem : CBVR
Previous efforts • Image feature vector indexing: • QBIC(1995), Photobook(1996), WBIIS(1997) etc. • Information loss : shape, location etc. • Image segmentation: • WindSurf(1999), Blobworld(1999), SIMPLIcity(2001) etc. • No iterative refinement
An interactive CBIR system • User enters a query image • System • Quickly segments the query image • Searches images with similar segments • Returns the approximate results • User iteratively refines the results A progressive refinement strategy • Our focus : Quick segmentation
Underlying requirements • Hierarchical image segmentation • Control over levels of segmentation • Fast segmentation for quick response • Intuitive segmentation
Normalized cut (Ncut, Shi et al. ‘00) Global Optimization Good criteria Promising in videos Cost Time : O(n1.5) Space : O(n2) 106 pixels or 200x100, 50 frames video: time : 1,000,000,000units, space : 1,000,000,000,000units Hierarchical segmentation Image with 2 segments
We need … • Speed + Quality How ? Reduce input size fed to algorithm ! Input size : n1/2, Cost : O((n1/2)1.5)
Two step pipeline Hierarchical segmentation Fast grouping O(n log n) n=9x104 Ncut O(n1.5)
Fast grouping • Group similar pixels • Grouping using local variation (PAlgo Pedro et al. 1998) • Uses local properties • Fast : O(n log n) • Produces too many regions for CBIR
Pipeline • i/p PAlgo intermediate Ncut o/p • Pipelining not easy • Unpredictable output size of PAlgo • Needed output size : n1/2 • Keep the quality intact • Solution • Cluster merging
Merging process • Sort the ‘similarity’ • Merge ‘similar’ groups iteratively
Result Pipeline O(n log n) Input N-cut O(n1.5)
Conclusion • A new pipelining strategy • Efficiently combined two approaches • Application in CBIR • Possibilities • Video segmentation • Video retrieval system (CBVR)
Thanks!! Questions ? aranjan@dgp.toronto.edu http://www.dgp.toronto.edu/~aranjan
Why N-cut ? • Edge weights are proportional to similarity