The microstructure of MoSiBTiC alloys is very complex, with three to four constituent phases and characteristic structures such as fine precipitates and lamellar structures. To perform the microstructural analysis efficiently, image segmentation was first performed for each phase of the microstructural images. Utilizing the Trainable Weka Segmentation method based on machine learning, the required segmentation time was dramatically reduced. Furthermore, by pre-adjusting the contrast of the images, the segmentation could be performed accurately for gray phases with different shades of gray. In addition, the U-Net method, based on deep learning, could perform highly accurate segmentation of characteristic microstructures consisting of multiple phases. The correlations between microstructural features and hardness were investigated using the segmented images in this study. The findings revealed that the volume fraction of each phase and the number of TiC clusters within the field of view significantly influenced hardness. This suggests that the hardness of MoSiBTiC alloys may be controlled by controlling the amount of TiC precipitates.