Analysis of Pulmonary Tuberculosis depends on the detection of non-specific and subtle variations in radiographic images. World Health Organization highly recommends Chest Radiographs (CXR) for the preliminary investigation and early diagnosis of TB due to its portability and non-invasive characteristics. However, computerized diagnosis of TB CXRs is challenging due to localized variations in lung texture and geometry producing large intra and inter image variabilities. In addition, it is difficult to identify non-specific abnormal patterns in diseased X-ray images. Hence, a pipeline of processes has been performed in this research work to characterize and differentiate normal and TB subjects in segmented chest radiographic images using local texture descriptors. The methods, significant results and the salient contributions of this research would be presented in this talk.