Photon-counting-detector (PCD) CT can provide multiple energy bin data sets and allows single-acquisition, multiple-contrast-injection imaging using materials such as iodine, gadolinium and bismuth. However, due to technical limitations, PCDs can suffer from compromised energy-resolving capability, which degrades multicontrast imaging performance. In this work, we investigate the use of a dual-source (DS)-PCD system architecture with additional beam filtration to improve spectral separation among energy bin data sets, and quantify its performance for multi-contrast imaging. Experiments were performed using a CT phantom including various concentrations of iodine (I), gadolinium (Gd) and bismuth (Bi). The DS-PCD architecture was emulated by scanning the same phantom twice on a single-source (SS) PCD-CT with two different tube potentials: 80 kV (energy thresholds = 25/50 keV), and 140 kV (energy thresholds = 25/90 keV) with a 0.4-mm tin filter. We further compared material decomposition performance using the proposed DS-PCD approach with that of the current SS-PCD approach. For the SS-PCD, chess mode with 4 energy bins was used, with energy thresholds of 25/50/75/90 keV to resolve the K-edges of Gd and Bi. The mean energies of the four energy bins in SS-PCD were 72/76/93/109 keV, while those of the four energy bins using DS-PCD were 57/64/88/111 keV, denoting a better spectral separation using DS-PCD. The material quantification root mean square error (RMSE) was reduced from 4.5/3.3/1.2 mg/mL for iodine/Gd/Bi using SS-PCD, to 1.4/1.2/1.1 mg/mL using DS-PCD. These results demonstrate that the DS-PCD can improve multi-contrast imaging performance compared to a SS-PCD acquisition.