Content analysis in media mobile retrieval

• Main Author: Liang, Wei Keong.
• Other Authors: Yap Kim Hui
• Format: Final Year Project
• Language: English
• Published: 2010
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Wireless communication systems
• Online Access: http://hdl.handle.net/10356/40191
• Institution: Nanyang Technological University

In recent decades, mobile phone industry is undergoing tremendous changes in areas of communication and multimedia. Furthermore, with the increase demand in generation of digital media through capturing and storing images in media mobile storage, there is a need for content management system to provide a retrieval of digital media based on users‟ request. Content-Based Image Retrieval (CBIR) is a methodology used in the process of image retrieval. In a CBIR system, an image can be represented by their own visual features such as color, shape and texture instead of the conventional method of representation by text-based keywords. Relevance feedback can be introduced to the CBIR system in providing interaction between user and the system to create more accurate results. In the first part of this project, modification to the original CBIR was done by increasing the number of images retrieved and thereafter sent for feedback. The performance efficiency increase is 0.44%. As the performance result of the modified CBIR system is not satisfactory, further improvement to the original system was carried out by integrating saliency technique onto the CBIR system. In the implementation of our saliency algorithm, mapping or detection of salient region in the images are first done before presenting to the Support Vector Machine (SVM) for classification. After executing the saliency algorithm, an additional step of defining a boundary region to capture the most salient region was also implemented. This is done so before integrating the saliency and CBIR algorithm. This is to ensure that the most salient region is send to the retrieval system for classification. A performance evaluation was then conducted to determine the effect and relevance of the saliency algorithm on our CBIR system. The experimental results have shown that the performance of the CBIR system when integrated with the saliency algorithm has increased by 1.46% as compared to the CBIR system without using the saliency technique.