MOESM5 of Image-based quantitative analysis of tear film lipid layer thickness for meibomian gland evaluation HwangHyeonha JeonHee-Jae YowKin HwangHo ChungEuiHeon 2017 Additional file 5. A figure that shows how to filter out frames that include eye-closing motion or closed eye. We set the value to less than the average between $$ {\text{B}}_{\text{opened}} $$ B opened and $$ {\text{B}}_{\text{closed}} $$ B closed (threshold = 0.5) to prevent the transitioning frames be regarded as falsely opened state (ROI data). On the other hand, we had to set the threshold higher than 0.25 which was the variation of ROI brightness. Thus, we set our threshold as 0.33 which provided reasonable results most of the time (98% of the case).