We can better encode
the ROI (or background image) at the network level coupling a
descriptor with each fragment and enhance its degree of uncorrelation.
In other words we can define a ‘‘self-reconstructing’’ packet
where the header contains all the information required for reconstructing
the bytes included in the local payload (see Fig. 9). The
following fields are contained in the header format: ID is the image
identifier (to keep transmitter and receiver synchronized); x and y
are the absolute coordinates of the first pixel in the payload (in the
reference frame of the image); wROI represents the ROI (or background
image) width and wr is the residual width representing
the remaining bytes to be written in the y row. This encoding technique
introduces a certain transmission overhead thus worsening
the compression ratio and, consequently, the bandwidth used to
send an image. On the other hand, it permits a stronger error-resilience
because of the higher degree of uncorrelation and the granularity
(self consistency) introduced at the fragment level. In order
to avoid mixture of ROIs in fragments, in our simulations we padded
by zeros the final fragment referred to a ROI.