1 Visual Information Extraction in Content-based Image Retrieval System Presented by: Mian Huang Weichuan Dong Apr 29, 2004
2 Talk Outline Introduction The Image Retrieval System Visual Feature Extraction The Experiments Discussion and Future Work
3 Introduction Motivation Efficient Image Retrieval – large, varied digital collections Images from
4 Introduction Background Text-based Image Retrieval Drawback Content-based Image Retrieval Visual content/feature Color – our focus Texture Shape Position
5 The Image Retrieval System Content-based Image Retrieval System Architecture X. Li, S. Chen, M. Shyu, B. Furht, Florida International University, Miami
6 The Image Retrieval System(cont.) The role of Visual Information Extraction Image DBFeature DB Color Label Histogram Computation Segmentation Algorithm Visual Content Extraction Feature Indexing Feature Indexes here we are
7 Visual Information Extraction The Algorithm Group - Color label histogram computation Choosing color space - HSV Categorizing colors - bins Describe - Image labeling Work on Colors – why color histogram Robust – scaling, orientation, perspective and occlusion Joint distribution of 3 color channels – global info. Fast
8 The Experiments Expected results Input Output Images from
9 Future Work Improved segmentation algorithm Color histogram – not sufficient Spatial information of color Joint distribution of color, texture and spatial features Test on images Distinctive objects Distinctive scenes Distinctive objects and scenes other