The delineation of liver tumors from computed tomography (CT) scans holds considerable significance in computer-assisted diagnosis, surgical strategizing, and treatment supervision. Nonetheless, achieving precise and resilient tumor segmentation poses a formidable challenge. This challenge stems from the indistinct boundaries and subdued contrasts between the tumors and adjacent tissues, as well as the considerable diversity in tumor characteristics across patients, including intensity, morphology, and location of the tumors. In this paper, we propose a robust approach for liver tumor segmentation utilizing marker-controlled watershed transform and make a comparative analysis of the segmentation outcomes with adaptive region growing and graph cuts segmentation algorithm. Initially, the tumors’ rough segmentation and the corresponding regions of interest (ROIs) are derived. Subsequently, the ROIs undergo refinement via median filter, leveraging the intensity distributions of the initially segmented tumor regions. Finally, the amalgamation of this refined information with gradient data is employed to accurately and efficiently extract the tumors from the ROIs. This method demonstrates resilience to noise, circumvents the need for pre-segmentation of the liver, and sidesteps the intricate and laborious training procedures. The segmentation is performed on real time datasets of liver tumor captured in Primus diagnostics Centre, Guwahati, Assam.