An adaptive algorithm for extracting foreground objects from background in
videophone or videoconference applications is presented in this paper. The
algorithm uses a neural network architecture that classifies the video fram
es in regions-of-interest (ROI) and non-ROI areas, also being able to autom
atically adapt its performance to scene changes, The algorithm is incorpora
ted in motion-compensated discrete cosine transform (MC-DCT)-based coding s
chemes, allocating more bits to ROI than to non-ROI areas, Simulation resul
ts are presented, using the Claire and Trevor sequences, which show reconst
ructed images of better quality, as well as signal-to-noise ratio improveme
nts of about 1.4 dB, compared to those achieved by standard MC-DCT encoders
.