Image Enhancement Using Color Space Components for Effective Tuberculosis Detection

Tuberculosis (TB) is an infectious disease caused by Mycobacterium tuberculosis bacteria. Image enhancement in ZN-stained sputum smear microscopy (SSM) images is an essential technique involved in automatic TB detection to improve image perception. However, only a few studies on TB diagnosis have been published in high-impact journals, and the majority of them used an ineffective sputum smear image enhancement technique, resulting in unsatisfactory segmentation results. Hence, we propose a novel image enhancement method, namely 'sputum smear image enhancement using effective components of color spaces (SSE-CCS)' for ZN-stained SSM images. This approach employs an iterative mean filter to suppress impulsive noise and the effective enhanced components of color spaces such as RGB, HSV, and NTSC/YIQ in order to improve visibility on both overlapped and non-overlapped TB bacilli regions. Also, two new powerful preprocessing algorithms, namely 'enhancement in non-overlapped bacilli region using V, Gray, and Q components (ENOB-VGQ)’ and 'Enhancement in overlapped bacilli region using S, Q, and Green components (EOB-SQG)' is proposed to improve the object clarity in both the TB bacilli regions. Experiment results show that the proposed SSE-CCS algorithm improves the mean-difference metric up to 5.64 times, indicating that it is well suited for the tuberculosis bacilli segmentation process. Furthermore, as compared to other existing image enhancement algorithms, it provides additional information about the images, which helps pathologists to detect the TB bacteria more correctly.