Multimodal Sensing Interface for Haptic Interaction

Figure 20

Object description algorithm. The color image (a), presented to illustrate one sample object, is not part of the processing. The mask of connected components (b) obtained in the former stage is the initial input of this block used. The depth pixels corresponding to each object are segmented using the mask (c). Histogram equalization (d) is used to find the brightest pixel in the image (e). The brightest pixels in the object correspond to the corner (or edge) closer to the Kinect-I. This corner is used to describe the height of the object. The contour in (c) is filtered to find other points with a similar height to the corner (f). The “corner” point and the points with similar height are used to describe the top shape of the sample.

(a)

(b)

(c)

(d)

(e)

(f)