Method for Medical Image Retrieval Based on Frequency Layer Feature

Abstract

Medical image features extraction is a crucial part in medical image retrieval system. In order to extract the medical image feature more accurately for improving the retrieval efficiency, a sort of novel feature called ”frequency layer feature ” and a method used for extracting the feature were proposed in this paper. First, the image was decomposed several sub-images in different frequency layer according to the characteristics of medical image. Then color features based on 2D color-order graphs and shape features based on Moment Fourier Descriptor (MFD) for the sub-images were extracted. At last, PCA method was used for compressing feature dimensions to reduce computational load. Experimental results shows that the novel method proposed improved evidently precision and recall rates of medical image retrieval than the traditional algorithm.