Medical Image Retrieval Using Local Binary Patterns with Image Euclidean Distance

Abstract

For years, researchers in medical image retrieval area have been representing and recognizing medical images based on local binary patterns (LBP). Compared to Gabor wavelets, the LBP features can be extracted faster in a single scan through the raw image and lie in a lower dimensional space, whilst still retaining image information efficiently. To improve the recognition rate, several methods using local binary pattern (LBP) have been tried such as improved local binary pattern (ILBP), extended local binary pattern (ELBP) , extended local binary pattern (ELBP), local Gabor binary pattern (LGBP). This paper proposes a novel medical image retrieval method, local binary pattern with image Euclidean distance (IMED), which takes into account the spatial relationships of pixels, and it is robust to small perturbation of images. Experiments showed that IMED improved the performance of standard LBP algorithm.