Abstract:

In allusion to SVM,s defects of handling large amount of data and distinguishing the importance of the training set,this paper joins the SVM classifier with the rough sets theory, and constructs an improved image feedback retrieval algorithm based on rough sets and SVMs,which are used to retrieve thyroid CT images .The results show that the improved SVM classifier can get 92.53% accuracy which is about 15.95% higher than 76.58% using SVM,and the retrieval of poor accuracy and recallprecision are also increasedr by 89.53% and 29.67%.

Key words: image retrieval;rough sets;SVM;relevance feedback, thyroid CT images