Abstract
A comprehensive survey on patch recognition, which is a crucial part of content-based image retrieval (CBIR), is presented. CBIR can be viewed as a methodology in which three correlated modules including patch sampling, characterizing, and recognizing are employed. This paper aims to evaluate meaningful models for one of the most challenging problems in image understanding, specifically, for the effective and efficient mapping between image visual features and high-level semantic concepts. To achieve this, the latest classification, clustering, and interactive methods have been meticulously discussed. Finally, several recommendations for future research issues have been suggested based on the weaknesses of recent technologies.
| Original language | English |
|---|---|
| Pages | 775-779 |
| Number of pages | 5 |
| Publication status | Published - 2010 |
Keywords
- content-based retrieval
- image classification
- image retrieval
- image sampling
- pattern clustering
- content-based image retrieval
- patch recognition
- patch sampling
- image visual feature
- high-level semantic concept
- image clustering
- interactive method
- Image segmentation
- Testing
- Databases
- Fires
- Review
- content-based image retrieval (CBIR)
- semantic concept
- effective mapping
- patch characterizing
- patch recognizing