Automatic Image Annotation using 2D MHMM

Abstract

Automated Image Annotation is an important and challenging task in the field of computer vision and CBIR(Content-Based Image Retrieval). It has extensive use in research as well as personal fields. In this project, the same has been achieved with the help of a statistical method, namely a 2-dimensional multi-resolution hidden Markov model. Prior to classifying images by the system, it is trained using a set of images which are previously annotated using labels. Then the image to be annotated is compared against each trained model produced as a result of the previous step. This produces a parameter called likelihood. The label having the highest likelihood is assigned to the image.