130 likes | 254 Vues
This paper explores the Query-by-Concept (QBC) approach for improving image retrieval systems. It discusses the necessity of advanced image retrieval techniques and contrasts various methods, including Query-by-Example (QBE). The Automated Sampling Image Annotation (ASIA) method utilizes a monotonic tree structure to organize images in a hierarchical system, enabling effective feature extraction and concept-based querying. The study demonstrates how QBC enhances image annotation through high-level concepts, improving retrieval speed and accuracy in complex images, with implementation case studies and experimental results.
E N D
Annotation techniques for Query-By-Concept Approach in Image Retrieval System Rakesh Kamatham Venkata
Introduction • Need for Image Retreival System • Different Approaches • Query-by-Example(QBE) • Query-by-Concept(QBC)
Techniques for QBC • Monotonic Tree • ASIA: Automated Sampling-Image Annotation
Monotonic Approach • Organizes the image in Heirarchial Structure
Components of the system • Image Database • Feature Extraction • Image Querying
Case Studies: • Building
ASIA Technique • Annotation is done by high level concepts such as ‘car’, ‘water’ etc. • Datbase images are matched based on the concept.
Conclusion • Monotonic approach annotates the images based on a model. The model does not differentiate the real world objects. • AISA approach can annotate the images which are complex. It does the matching offline and is much faster in retrieving the image.