ROIs: change the internal data model

[PierreRaybaut]

Currently, the ROI data model is simplistic:

• It stores all the ROI associated to a signal or an image in an integer NumPy array
• For images, N rectangular and circular ROIs are serialized in a N x 4 array:
  • Each line of the array are coordinates: x0, y0, x1, y1
  • For rectangular ROIs, each line of the array simply contains the top-left (x0,y0) and bottom-right (x1,y1) coordinates
  • For circular ROIs, the four coordinates are the X-diameter coordinates (y0=y1)

The idea of this Issue would be to modify this model in order to be able to store:

• Polygon ROIs (for images)
• A parameter for each ROI: inside (current behavior) or outside - meaning that the ROI is inside or outside the defined area

The new internal data model could be a dictionary:

• for signals:
  • keys: "segment" (the only key...)
  • value associated to "segment" would be a dictionary with the following keys:
    • "coords" (the NumPy array, identical to current roidata attribute)
    • "inside" (a 1D boolean NumPy array)
• for images:
  • keys: "rectangle", "circle" (in the near future: "polygon")
  • values associated to each key:
    • "coords" (the NumPy array, identical to current roidata attribute)
    • "inside" (a 1D boolean NumPy array)