Neighborhood sequences for comparing similarity vectors in image retrieval

In this paper, we present a novel approach for image database retrieval, which is a developing field with growing interest. Many features (e.g. color, shape, texture) can be considered to find similar images in the database for a given query image. We recommend a novel way of measuring the similarity which is based on calculating the norm of the similarity vectors (that is the distance between the query image and the other images in the database). Our approach is based on neighborhood sequences to calculate this norm, because via neighborhood sequences we can give more flexibility in the composing the queries than the classic norm-computing methods.