Mobile media annotation, search and retrieval

• Main Author: Chew, Siew Mooi.
• Other Authors: Yap Kim Hui
• Format: Final Year Project
• Language: English
• Published: 2011
• Subjects: DRNTU::Engineering::Computer science and engineering::Information systems::Information storage and retrieval
• Online Access: http://hdl.handle.net/10356/45802
• Institution: Nanyang Technological University

Image annotation has been extensively researched upon, and is an important influence in image search and retrieval. This final year project aims to investigate different approaches to image annotation, by contextual information annotation and content-based image retrieval. For the first aspect, the Gaussian Mixture Models have been used effectively to represent keywords and time information of the images in the database, and could be used in exploring further on integration of contextual information to create an accurate automated image annotation process. For the aspect on content-based search, the effects of scene recognition using color SIFT descriptors were investigated. Color SIFT descriptors have been used for local feature representation after the Harris Laplace salient point or dense sampling detectors have been executed for feature extraction. The Bag of Words framework is adapted, and each image is represented by a set of local features plotted over a histogram. The hierarchical k-means approach is used in clustering, and the Support Vector Machine is used in machine learning. These methods were selected from previous works, which proved their robustness as compared to other techniques. From the testing results, it is proven from our database of scene categories that the descriptors OpponentSIFT and CSIFT performed better, and the dense sampling approach also yielded better results.