Abstract: Seismic tomography provides a fundamental approach for probing the internal structure of the Earth. In recent years, imaging methods based on shear-wave splitting intensity (SI) have steadily advanced, enhancing investigations into upper-mantle anisotropy. In practice, SI is measured by projecting the transverse component of a shear wave onto the time derivative of the radial component. Under low-frequency conditions and for small delay-time conditions ( \omega \textδt\ll 1\textor\omega \textδt\rightarrow 0 ), SI follows a sinusoidal relationship with the splitting parameters, \varphi and \textδt . By retaining the complete mathematical expressions of the radial and transverse components and substituting them into the projection-based integral definition, a new relationship incorporating the dominant angular frequency, \omega , can be derived. While, current anisotropic imaging studies generally adopt waveform-derived SI as the observational constraint, the imaging performance of SI calculated from pre-estimated splitting parameters remains poorly understood. Focusing on the Sichuan-Yunnan region, this study systematically compares anisotropic imaging results obtained from three types of SI: waveform-measured SI, SI calculated using the conventional sinusoidal relation, and SI derived from the new frequency-dependent formulation. The results demonstrate that, compared with the conventional sinusoidal formulation, imaging based on the new relationship yields structures that are more consistent with those obtained from waveform-derived SI, indicating its improved reliability for anisotropic tomography.