Steerable Texture Descriptor for an Effective Content-Based Medical Image Retrieval System Using PCA

Abstract

Digital images have increased in quantity especially in the medical field used for diagnostics. Content-Based Medical Image Retrieval System will retrieve similar medical images from large database based on their visual features like texture, color, and shape. This paper focuses a novel method to increase the performance using Boundary detection, Steerable filter, and Principal Component Analysis. The content of the image was extracted with the help of region-based texture descriptor using steerable decomposition followed by extracting Principle Component Analysis which has better feature representation capabilities. The similar medical images are retrieved by comparing the extracted feature vector of the given query image with the corresponding database feature vectors using Euclidian distance as a similarity measure. The effectiveness of the proposed method is evaluated and exhibited via various types of medical images. With the experimental results, it is obvious that the region-based feature extraction method outperforms the direct feature extraction-based image retrieval system.