Figure 2.1:
Analysis of a user in the ALIVE environment. The
frame on the left is the video input (n.b. color image shown
here in black and white for printing purposes), the center
frame shows the segmentation of the user into blobs, and the
frame on the right shows a model reconstructed from blob
statistics alone (with contour shape ignored).
Figure 2.2:
The user's hand and face blobs have
very similar color statistics. These two blobs end up very
close to each other in this frame.
The human is modeled as a connected set of blobs. Each blob has a spatial
(x,y) and color (Y,U,V) Gaussian distribution, and a support map
that indicates which pixels are members of the blob. We define 
to be the mean (x,y,Y,U,V) of blob k, and 
to be the
covariance of that blob's distribution. Because of their different
semantics, the spatial and color distributions are assumed to be
independent. That is, is block-diagonal, with uncoupled
spatial and spectral components.
Each blob has associated with it a support map, that indicates which
image pixels are members of a particular blob. We define
, the support map for blob k, to be
An aggregate support map s(x,y) over all the blob models is also a useful data
structure. Since the individual support maps indicate which image pixels are
members of that particular blob, the aggregate support map represents the
segmentation of the image into spatial/color classes.
Each blob can also have a detailed representation of its shape and appearance, modeled as differences from the underlying blob statistics. The ability to efficiently compute compact representations of people's appearance is useful for low-bandwidth applications, such as our demonstration of a shared virtual environments at SIGGRAPH '95 [6].
The statistics of each blob are recursively updated to combine information contained in the most recent measurements with knowledge contained in the current class statistics and the priors. Because the detailed dynamics of each blob are unknown, we use approximate models derived from experience with a wide range of users. For instance, blobs that are near the center of mass have substantial inertia, whereas blobs toward the extremities can move much faster.