Image segmentation for facial image coding of videophone sequences

A method to extract the facial area of videophone sequences is discussed. The proposed technique utilizes the distribution of skin colours in the colour space to obtain an initial segmentation followed by several successive post-processing operations to yield the final result. This approach is robust with regards to facial shape, size, orientation, or motion and can be implemented at a low computational complexity. The motivation is to enable a video coding scheme to place a greater emphasis on the facial region in a videophone sequence in order to produce a perceptually higher quality in the coded images. Experimental results are presented for a series of CIF format frames from two videophone sequences.