Reducing feature space dimensionality for image retrieval

In the field of image retrieval, images are represented using a feature vector of high dimensions, in the most of applications. Working with vectors of high dimensions consumes an enormous amount of memory and slows down the process. In this paper, we propose an algorithm for selection of some dimensions from the total set of feature dimensions which not only reduces the memory requirement and speeds up the retrieval process, but increases the precision of retrieval, whereas, in the literature it is perceived that reducing the feature space dimension reduces the precision.