Image Retrieval using Discrete Orthogonal Moments in a Non Uniform Lattice

Retrieval of images from large databases is an emerging application particularly in medical and forensic departments. Face image retrieval is still a challenging task since face images can vary considerably in terms of facial expressions, lighting conditions etc. In feature based image retrieval methods, the accuracy depends on the discrimination power of the features. In this work, orthogonal moments were employed as features for the retrieval task. Due to the orthogonal property, these moments are inherently non-redundant exhibiting good image representation capability. Racah moments defined in a non-uniform lattice are proved to be better than other orthogonal moments in terms of reconstruction error. Face image retrieval using Racah moment features has been extensively experimented with YALE face database and FERET Database. The results reveal the efficacy of orthogonal moment descriptors.