Axial T2 Weighted MR Brain Image Retrieval Using Moment Features

Abstract

AbstractMagnetic resonance images play a vital role in identifying various brain related problems. Some of the diseases of the brain show abnormalities predominately at a particular anatomical location which on MR appears at a slice at defined level. This paper proposes a novel technique to locate desired slice using Rotational, Scaling and Translational (RST) invariant features derived from a ternary encoded local binary pattern (LBP)image. The LBP image is obtained by labeling each pixel with a code of the texture primitive based on the local neighborhood. The ternary encoding on LBP identifies the boundary of the uniform region and thus reduces the time for calculating moments of different order. The distance function based on the RST features extracted from LBP between query and database image is used to retrieve similar images corresponds to the query image.KeywordsFeature ReductionRotational Scaling and Translational (RST) invariant featuresLocal Binary patternEccentricityPrecision & Recall