Certain analysis on content based medical image retrieval system using shape texture semantic and visual features

Abstract

Content-based image retrieval (CBIR) is a technique of searching the images based on the content of the image. CBIR is a promising technology to enrich the core functionality of picture archiving and communication systems (PACS) in medicine. CBIR in medical domain has a potential for making a strong impact in diagnostics, research and education. In this thesis, content based medical image retrieval systems based on new techniques have been proposed. In the conventional CBIR systems, the content of the image is represented by features such as color, shape and texture of the image. Various shape feature extraction methods are studied and a new shape feature extraction method is proposed. Various texture feature extraction methods are studied and a new texture descriptor is proposed. A medical image retrieval system based on this new texture descriptor is developed. The experimental results show that this new framework outperforms the other existing texture descriptor in medical CBIR systems. This research aims to investigate an effective CBIR framework for combining the primary image features. A multiple classifier framework for content based image retrieval in medical domain has been proposed and implemented by combining multiple shape and texture visual features. The deep gap between visual features and high-level semantics is a major obstacle to more effective image retrieval. To still improve the retrieval performance of the medical CBIR system, a new hybrid retrieval approach which combines visual shape, texture features and high level semantics is proposed. The advantage of this approach is that it supports both query by keyword and query by image content. From the retrieval experiments conducted, it is found that the proposed hybrid retrieval approach system has higher retrieval performance than the proposed multiple classifier system which uses only visual features. newline newline newline