A Generic Framework for Content based Image Retrieval and Similarity Re-ranking of Volume Rendered Stereoscopic 3D Medical Images

Abstract

Various 3D medical imaging systems have recently been developed to enable more accurate diagnosis in medical field. Content-based Image Retrieval (CBIR) has been identified as a key technology for computer-aided diagnosis. In this paper, we present a content-based image retrieval system intended to retrieve a volume rendered medical stereograms from the database based on visual cues such color, shape and homogeneous texture. The proposed scheme is query based, extract similar images from the database and re-rank the retrieved images based on the degree of relevancy. HSV histogram, Auto Color Correlogram and Color Moments are employed to extract color features whereas shape and texture features are extracted with Fourier descriptor and Weber Law Descriptor respectively. The similarity between the query image and retrieved images is measured with Mahalanobis distance measure. The disparity based re-ranking adopted for refining retrieved results. The experiments are conducted in a diverse collection of 815volume rendered medical stereograms of different modality. Different input queries applied to the proposed system and performance is evaluated based on the precision, recall and F1-score. The experimental result shows a promising performance with the multi-feature based image retrieval system for the volume rendered medical stereograms.