Feature-Based Similarity Retrieval in Content-Based Image Retrieval

Content-based image retrieval (CBIR), providing query by image examples other than key words, is a hot topic in recent years. Querying by words mainly depends on the performance of crawler, whereas query by example is more unpredictable, since feature extraction is still challenging due to the rich content of the image. This paper focuses on the issue of similarity retrieval in high-dimensional space, a problem of performing nearest neighbor queries efficiently and effectively over large high-dimensional databases. Although some arguments have advocated that nearest-neighbor queries do not even make sense for high-dimensional data, we review the existing techniques of working in vector space of high dimension, and provide our unique view towards the issue of time complexity and precision during similarity retrieval in CBIR.