Foreword
In RCSCUT it's possible to vectorize a skin contour from a picture loaded or from a photo directly shooted in the nesting table
But the old method was not entirely satisfactory and could be significantly improved
That's why a new one has been developped
Version
15.0.cp
Overview
New dialog
The dialog has been updated to allow better visibility of the different steps and parameters
The dialog is divided in 3 tabs
Pictures tab
This is the tab where you can see the reformed picture, either loaded from the disk or shoot from the nesting table
In this last case, you can set the retroprojection color light or the APN compensation during the shoot
Results tab
While the picture/photo is loaded, the skin contour is computed and displayed in this tab
It's displayed both on the dialog and on the nesting table with the retroprojection
It exists 3 methods to compute the skin contour
- Polarization: a new method based on a detection algorithm by color difference
- Threshold: a new method based on detection algorithm by threshold
- Legacy: the old algorithm used on the old dialog
Each method has its own parameters
The option Display picture allows to show the original picture with the skin contour detection
Enhance the detection: Reframe the useful zone
Sometimes the result could not be as good as expected by the user even adjusting the different parameters
Indeed there can be many reasons that the skin contour detection is not good
For example the luminosity, the closed colors between the skin and the background or also some artefacts in the picture
In these cases, it's possible to reframe the useful zone taken into account for the skin contour detection
Note: Only for the polarization and threshold methods
In this example, we use the polarization method but the skin contour is not found
Click on Reframe
From the application screen or from the retroprojection screen you can adjust the useful zone (cyan) by moving the four corners and frame the skin
Open the contextual menu with right click to
- validate
- set the default frame
- cancel the current modification
Now the result is good
The useful zone is represented by a cyan frame
Enhance the detection: Backgrounds tab
An other way to enhance the skin contour detection is to use a picture/photo of the background without the skin
The algorithm does a picture difference to find more easily the skin contour
An example with this skin
With the method Threshold, the skin contour is not computed
But with a background picture, the result is better