A motion based unsupervised neural network approach to motion segmentation is addressed, and embedded in an automatic object based coding system. The motion estimation phase is carried out by an arbitrarily shaped object oriented block based technique (S-BMA). An efficient polynomial motion model is used to describe motion fields and jointly segment images into background-foreground. The proposed technique is embedded in a H.263-like coding system and tested on the foreman sequence. Preliminary results on standard video sequence seem promising.
An Efficient Object Based Personal Video Coding System / Guaragnella, Cataldo; T., D'Orazio. - 3333:(2004), pp. 655-664. [10.1007/978-3-540-30543-9_82]
An Efficient Object Based Personal Video Coding System
GUARAGNELLA, Cataldo;
2004-01-01
Abstract
A motion based unsupervised neural network approach to motion segmentation is addressed, and embedded in an automatic object based coding system. The motion estimation phase is carried out by an arbitrarily shaped object oriented block based technique (S-BMA). An efficient polynomial motion model is used to describe motion fields and jointly segment images into background-foreground. The proposed technique is embedded in a H.263-like coding system and tested on the foreman sequence. Preliminary results on standard video sequence seem promising.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.