A survey on context based similarity techniques for image retrieval

Most of the current research efforts in multimedia domain have been done for improving the performance of the image retrieval system by reducing the semantic gap. Semantic gap is a gap between image low level features represented by machines and high level human perception used to perceive the image. Majority of the image retrieval system uses image low level feature such as content or visual properties of an image for retrieval. Image high level features that is image context such as keywords, tags, captions related to images are also important as they play key role in reducing the semantic gap more effectively when compared with image low level features. Therefore by considering image high level features in image retrieval systems we can easily reduce the semantic gap and effectively retrieve the images. This paper thus presents the survey of various techniques used for measuring the similarity between image contexts, their merits and limitations. From the surveyed techniques, we have also analyzed that most of the image context based similarity techniques requires very less computation cost using which an efficient image retrieval system can be built.