Video surveillance has taken a huge importance in the
everyday life, in order to detect, to recognize, to prevent
dangerous situations. This work aims to propose two
methods able to improve the robustness and speed of the
detection of moveable objects or pedestrian using a fixed
stereo camera. The first method is an improvement of a
real-time adaptive non-parametric threshold with a salt-andpepper
noise removal used to detect changes into images.
The second method instead proposes to improve the
detection of the shadow points in order to affine the
extraction of the pedestrians or objects from the scene.
Comparison results about accuracy and performance are
shown at the end of the paper.
