Remote Photoplethysmography (rPPG) realizes heart rate measure ment by capturing facial video and analyzing the periodic changes of the average gray level in the range of interest (ROI).Noise elimination is one of the key techniques in rPPG systems.Among all the noises,diffuse light on the skin surf ace is a recognized common noise.According to the principle of interaction betw een light and biological tissues,backscattered light without contact with subcu taneous artery should also be considered as noise,which is called ‘shallow noi se’ in this paper.According to the author′s literature research,there is no r esearch paper on this noise and its elimination method in the rPPG research fiel d.In this paper,the polarization characteristics of the imaging beam is system atically analyzed to improve the polarization based noise elimination technique, the results show that:(1) When the incident light is S- or P- wave,the pola ri zation direction of the backscattered and incident light is different.In this c ase,the diffuse light and shallow noise cannot be completely eliminated by maki ng the analyzer and polarizer orthogonal.The analyzer should be orthogonal to t he imaging beam; (2) The polarization degree of the imaging beam is not evenly d istributed.The polarization degree of the forehead is the highest,which can be used to optimize ROI selection.The results of this paper are helpful for rPPG systems based on polarization imaging to achieve better signal quality and impro ve the accuracy of heart rate measurement.