Comparative Analysis on Different Degrees of JPEG Compression Used in CBIR Systems

Content-based image retrieval systems (CBIR) use computer vision methods for organizing, analyzing, processing, and searching of million graphic files uploaded on the Internet daily. For that reason, the retrieval process is a key element of CBIR applications. The best system performance is based on accurate and complete search which must be finished in a reasonable amount of time. It is directly related to extracting, storing, and indexing all necessary metadata from the examined image set. This paper gives information about JPEG compression and how it is used in some of the most widespread CBIR systems. It reveals how different degrees of that compression influence the generated results from custom developed CBIR application. Furthermore, it shows various similarities between the same image set but examined in two different ways - with non-compressed and compressed data retrieved by the system. In the end, conclusions are made on the basis of the results obtained from developed and tested CBIR application which was made for these purposes.