A comparative study of local binary pattern descriptors and Gabor Filter for electron microscopy image segmentation

Abstract

This article focuses on the segmentation of mitochondria and synapses in electron microscopy (EM) images. Mitochondria and synapses are two prominent and important structures with neurological interest that play an important role in neural functionality. Manual segmentation of these structures in EM images by an expert neuroanatomist is a very tedious and error prone job. In this paper comparison of different feature sets based on Local Binary Patterns (LBP) and Gabor Filters for automatically performing this task are presented. In this study result analysis uses a random forest classifier and Jaccard index and ROC curve as comparative measures. A combination of Gabor filters and LBP features extracted from Gabor filtered images produce the best segmentation result.