A Comprehensive Study of Image Re-ranking Methods in Visual and Semantic Features

The objective of content-based image retrieval (CBIR) is to locate and retrieve similar images from the large-scale dataset for a given query image. A conventional CBIR system involves feature extraction, similarity measurements, and retrieval of the initial rank list. It is observed that, the images retrieved are visually dissimilar to the query images. The problem of research in CBIR is majority on improving the performance of the system whereby the query results are optimized to its maximum to receive the results with least number of dissimilar images. Toward the research efforts to improvise the performance, re-ranking strategies play crucial role. This paper offers a comprehensive examination of the re-ranking process over the last decade in CBIR. The categorization of this survey covers the different re-ranking algorithm on different retrieval lookout in improving the performance of object retrieval and object re-identification. This survey paper will lead in different directions regarding re-ranking image retrieval using other methods.