A Study On 2D Shape Analysis And Its Application In Image Retrieval

As a result of advances in the internet, multimedia technology and digital image sensor technology, the volume of digital images produced in everyday life and by educational, scientific, martial, industrial applications increased dramatically in recent years. The efficient management of the rapidly expanding images became an urgent problem. Content-based image retrieval , a technique which uses visual contents to search images from large scale image databases according to user's interests, then came to being and became an active and fast advancing research area.Images are the most important input in many machine vision systems, so the analysis of images is a key technology in these systems.The above issues form the background of this paper. The thesis focuses on 2D shape analysis technology and its application in content-based image retrieval.Many shape representation methods exist. Among them, methods based on Fourier descriptors (FD) achieve well representation. Different shape signatures have been exploited to derive FDs, however, FDs derived from different signatures can have significant different effect on the result of retrieval. The thesis makes a comparative study on Fourier descriptors with different shape signatures.Since Freeman chain code is a very popular and important method in shape description. The thesis combines chain code with Fourier series theory and proposes a new Fourier descriptor based on chain codes. Experiments show that this method is not suitable for image retrieval but has good performance in shape reconstruction.The thesis developed a CBIR prototype system named iSearch. The system provides sample based retrieval. Users are freely to select color, texture, shape features, or their combination in retrieval sessions. Some design patterns are deployed in the system design and development to make it expansible and reusable. Several retrieval experiments are conducted on iSearch and results are given in the thesis.