Reducing Semantic Gap in Image Retrieval by Integrating High Level Query and Low Level Facial Features

As a result of rapid increase in the digital images, annotation of human faces in images and retrieval of images is one of the desirable needs in the current world. For face annotation, face detection and recognition are two important tasks to be performed and is still challenging due to the wide variety of faces and the complexity of noises and image backgrounds. Automatic face annotation facilitates improved retrieval and organization of digital images. In the proposed work annotation of human faces is done using PCA and MLP. Two scenarios are for naming people in an image finding all faces, and assigning names to all faces. For naming, free text type of annotation is used. This makes annotation task easier, but more difficult to use the annotation later for image retrieval. Efficient image retrieval of an annotated face can be done using the combination of name label and Euclidean distance measurement mechanism.