Image Improved Annotation and Retrieval Algorithm Based on Relevance Feedback

摘要

In order to improve the performance of image retrieval, an image annotation and retrieval algorithm is proposed on the basis of segmentation and relevance feedback. The algorithm utilizes the correlation between visual features and tagging words, adopts image visual characters of regions, and obtains a group of visually similar images by clustering. Then it not only take such image features as color, shape into consideration, but calculates the similarities between the region and the nearest three classifications, and integrates the keyword probability vector(KPV) to obtain the most appropriate KPV. The proposed algorithm also employs user's feedback information to adjust the relationship between the query words and each classification, and to improve accuracy of the image retrieval. The experimental results show that the proposed algorithm betters precision and recall of image retrieval.