Refining

The aim of this step is to analyse the output for a given and be able to determine how good the output is, in order to be more accurate for the next frame's process.
Cause -> Consequence

Relative spatial positions -> Adjust Boxes positions
Size of features -> Scale Boxes
Number of blobs inside boxes -> Change structural kernel for Dilations
Area of detected blobs -> Increase or decrease Threshold values

**Figure 1:** Output image (left) and refined image

Index | Facial Features Detection | Feature Points Extraction | Refining | Tracking