This paper presents a groupwise graph-theory-based parcellation approach for defining nodes in network analysis using resting-state fMRI data. The method aims to identify functionally homogeneous subunits in the brain, which can serve as a functional atlas for fMRI. The approach involves a groupwise optimization that ensures functional homogeneity within each subunit and consistency at the group level. Parcellation reproducibility across multiple groups of healthy volunteers is high, and the method is shown to be effective for defining 100, 200, and 300 subunits. The results demonstrate that the proposed approach outperforms existing methods in terms of classification accuracy and spatial continuity. The method is applied to resting-state fMRI data from 79 healthy subjects, and the results show high reproducibility and spatial coherence. The parcellation results are available online along with tools for interfacing with analysis packages. The study highlights the importance of defining functionally homogeneous subunits for accurate network analysis and provides a novel approach for whole-brain parcellation based on graph theory. The method is validated using synthetic data and real fMRI data, showing its effectiveness in producing accurate and spatially homogeneous parcellations. The results also indicate that the proposed approach is more reliable than existing methods in terms of reproducibility and spatial continuity. The study contributes to the field of functional MRI by providing a new method for whole-brain parcellation that can be used for both connectivity studies and task-based activation studies.This paper presents a groupwise graph-theory-based parcellation approach for defining nodes in network analysis using resting-state fMRI data. The method aims to identify functionally homogeneous subunits in the brain, which can serve as a functional atlas for fMRI. The approach involves a groupwise optimization that ensures functional homogeneity within each subunit and consistency at the group level. Parcellation reproducibility across multiple groups of healthy volunteers is high, and the method is shown to be effective for defining 100, 200, and 300 subunits. The results demonstrate that the proposed approach outperforms existing methods in terms of classification accuracy and spatial continuity. The method is applied to resting-state fMRI data from 79 healthy subjects, and the results show high reproducibility and spatial coherence. The parcellation results are available online along with tools for interfacing with analysis packages. The study highlights the importance of defining functionally homogeneous subunits for accurate network analysis and provides a novel approach for whole-brain parcellation based on graph theory. The method is validated using synthetic data and real fMRI data, showing its effectiveness in producing accurate and spatially homogeneous parcellations. The results also indicate that the proposed approach is more reliable than existing methods in terms of reproducibility and spatial continuity. The study contributes to the field of functional MRI by providing a new method for whole-brain parcellation that can be used for both connectivity studies and task-based activation studies.