A Review on Content Based Image Retrieval using Mixture Model for Image Representation

Content Based Image Retrieval has drawn attention in last few years. CBIR is used for retrieving images from the large scale. Without the use of compressed code for image retrieval large scale of images cannot be searched expeditiously. Bag-of-Words BoW is not realistic for huge image database because it is concentrated in memory. The main objective is to achieve high performance and lower memory cost. The new framework Mixture of Subspace Image Representation is used. In this local descriptor is modelled for each image present in the database. A query image is given to the system is matched with the image present in the database and the likelihood function is computed. It is found out through the review of papers that the new framework represents every database image in a small amount of number of bytes which achieves good performance than the state-of �art. The memory required to store the data is more so the new framework is proposed. So that they require the less memory and improved performance of retrieval. The retrieval performance is improved because of the reduction of the large scale database.