Abstract
With the emergence of multimedia databases, exact keyword search performed in traditional databases is not applicable due to the complex semantic nature of multimedia data. In this paper, Content Based Image Retrieval approach was introduced to solve this problem by providing metadata for multimedia databases based on their actual contents (features) rather than raw keywords description. The search is based on similarity matching rather than exact match because of the fact that images are rarely identical. The used images in the experiments were obtained from Grimace facial images dataset available from the University of Essex, England. Similarity between database objects (images) was calculated using Euclidean, City-block and Chi-square distance functions. The most attractive results of the conducted experiments were obtained using City-block and Euclidean distance functions. Image’s features that can perform well when used individually were identified. Features that can perform well when combined with other features were also identified, in addition to excluding features that have limitations in distinguishing images such as image entropy value.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have