This work aims at detecting cardiac cavities in
2D echocardiographic sequences. In this respect, we propose
to extend to the case of moving images, a competitive fully
automatic intraframe segmentation method. The latter handles
the spatial information by characterizing the texture in each
image and, involves the watershed algorithm using an optimal
texture gradient image and a suitable seeds image. The novelty
of our work consists in accounting for the dynamic evolution
according to two strategies. The first one consists in projecting
and refining the boundaries from a frame to the next one. The
second strategy acts on the inputs of the watershed algorithm.
Experiments are carried out on apical 2D-echocardiographic
sequences and, they show the benefit that can be drawn from the
proposed strategies in terms of both objective accuracy metrics
and computational complexity.