A Review on Multi-Attribute Assisted Reranking using SVM Classification for Web Image Search

This review paper formulates an image reranking problem to improve Text Based Information Retrieval (TBIR) by using Multi-attribute learning methods. Existing methods train separate classifier for each word and heuristically combine outputs for retrieving multiword query. The proposed work partition relevant images into clusters, known as container, by using visual and textual attributes. Based on multi-attributes learning we cluster multiple relevant images in positive container whereas few relevant images are clustered with irrelevant images in negative container. Further we enhance multi-attribute learning algorithm to effectively rerank relevant images.