Sketch-Coding of Image Sequences for Video Conferencing

We consider the problem of obtaining a sketch of a sequence of grey level images and coding it at very low bit rates. The sketch is either binary or has very few levels (3 or 4). We use an image segmentation algorithm, presented in a previous paper, to obtain the sketch. It is an adaptive thresholding scheme that uses spatial constraints and takes into consideration the local intensity characteristics of the image. We have also added temporal constraints to the algorithm to ensure that motion is smooth. The algorithm is applied to images of human faces. In particular, we are interested in head and shoulders images. The sketches preserve most of the information necessary for recognition. They are similar to what an artist would sketch with a few brushstrokes. The movement of the face and in particular of the eyes and lips is rendered very accurately.