Recently, researchers have discovered that there are non-invasive ways to detect Diabetes Mellitus based on the analysis of human facial blocks. Even though a few algorithms have been developed to detect Diabetes Mellitus based on facial feature analysis, Diabetes Mellitus detection remains a challenging problem. In this paper, we propose two algorithms to detect Diabetes Mellitus through the analysis of facial texture features extracted by the Gabor filter: Simplified Patch Ordering and Improved Patch Ordering. Firstly, four facial blocks are taken from a facial image to represent it. Afterwards, we use a 2-D Gabor filter bank to extract texture features from the four facial blocks. Finally, Simplified Patch Ordering and Improved Patch Ordering are applied to classify Diabetes Mellitus and Healthy samples. Experimental results on a dataset show that Simplified Patch Ordering can classify Diabetes Mellitus and Healthy samples with an accuracy of 95.83%, a sensitivity of 100%, and a specificity of 91.67%, while Improved Patch Ordering can classify Diabetes Mellitus and Healthy samples with an accuracy of 99.38%, a sensitivity of 100%, and a specificity of 97.35%, both based on a combination of facial blocks.