Rapid increase in the use of digital images either for security, health treatment, or entertainment demands an effective image retrieval system. In text based image retrieval, images are annotated with keywords based on human perception. A user query is composed of keywords according to his/her requirements. Query keywords are matched with the keywords associated with images, for retrieval. This process has been extended with ontology to resolve semantic heterogeneities in keywords matching. However, crisp annotation and retrieval processes could not produce the desired results because both processes involve human perception. To reduce the matching complexities produced due to human perception, an image retrieval model has been proposed in this thesis that makes use of fuzzy ontology for improving retrieval performance. For representing the image content, it is divided into regions in our dataset and then regions are classified into concepts. The concepts are combined into categories. The concepts, categories and images are linked among themselves with fuzzy values in ontology. The model has been evaluated through both objective and subjective measures. Experimental results show that the proposed system performs better than the existing systems in terms of retrieval performance. Besides this, users usually desire higher proportion of the query keywords in the retrieved images than other undesired keywords. Existing systems return images that mostly do not contain the query keywords either in equal or higher proportion than other keywords. The research in this thesis resolves this issue by applying uncertain frequent pattern mining on the association that exists among the concepts in images. These patterns assist in retrieving images that contain the required query keywords in high proportion. The ranking of retrieved images has been objectively evaluated using two different measures. Experimental results show that the proposed image retrieval system performs better than existing image retrieval systems. The results of image retrieval systems are typically assessed for quality either by objective or subjective evaluation measures. These measures usually produce good results in typical image collections with predefined categories. However, the results of objective evaluation measures deteriorate in collections where an individual image may belong to multiple categories. Furthermore, conducting subjective evaluation is very difficult if not impossible on a large number of queries for every collection due to humans’ involvement, as it is a tedious and time consuming task. Therefore, an automated assessment model for subjective evaluation in image retrieval systems is required. The main hurdle in creating such an automated system is the availability of subjective evaluation benchmark for the retrieved images. This thesis also presents a new benchmark and a novel evaluation model for conducting automated subjective evaluation by tackling it as a supervised machine learning problem with support vector regression (SVR). The experimental results demonstrate that the proposed system automatically predicts the mean opinion score (MOS) with reduced error and correlates well with human subjects’ assessment.