Semantic Features Selection and Representation for Facial Image Retrieval System

Most image retrieval methods are specialized for image based matching and retrieval based on low-level features. However, humans perceive facial images and compare their similarities using high-level features, such as the description and ranking of facial traits including gender, race, and age. In this research, a new proposed method is used for semantic features selection and representation to be described by the user directly through appropriate verbal descriptions using the natural language concepts. In this manner, the research has strategized to bridge the semantic gap between the low-level image features and high-level semantic concepts. A total of 100 respondents participated in selecting and weighting the semantic features of the human face based on the level of importance of each feature. The semantic features were also integrated directly with the eigenfaces and color histogram features for facial image searching and retrieval to enhance retrieval accuracy for the user. Euclidean distance was used for features classes integration and classification purposes. The proposed human facial image retrieval has been evaluated through several experiments using the precision and recall methods. The results have indicated high accuracies, which are considered a significant improvement compared with low-level features based facial image retrieval techniques.