Multimedia Search Reranking

Image retrieval is the general term for the technique of searching images from a huge collection of image database and returning to the user all the images that matches the user query. The use of digital image databases in various fields like advertising, medical, commerce, education, industry and personal album in social media is rapidly growing, thus the image retrieval becomes a highly active research area for searching these digital images. This area is gaining more and more importance with the immense growth of multimedia content over the internet. The two main categories of image retrieval are the textual based image retrieval and the content based image indexing and retrieval. In textual based retrieval, textual annotations, character and numeric denotations are used to describe images and are submitted to the search engine as a query. Even though keyword based image retrieval displays most of the relevant images related to the query, it has some disadvantages. Firstly, annotating the image database is labour intensive. Secondly, each individual perceive and annotate images differently and the retrieval is not based on the actual content of the images. Hence it may retrieve some irrelevant (noisy) images which are also related to the text in some way. The content based approach address the shortcomings of the text based technique.