Accurate automatic frame fitting for semantic-based moving image coding using a facial code-book

A method of automatic wire-frame fitting for semantic-based moving image coding is proposed. The algorithm describes a technique for detailed fitting of the wire-frame in the areas occupied by important facial features (lips, eyes, eyebrows, nose). The algorithm utilises a preprocessed data-base of facial images. Only the monochrome edge maps are used in the automatic wire-frame fitting algorithm. This reduces the space required for storage of the reference data-base. The images from the data-base are pre-processed and manually fitted with the wire-frame model as part of the data-base preparation. The processing of an unknown image is therefore accomplished in a fast and efficient manner. A new measure of the accuracy of the automatic wire-frame fitting is proposed. Encouraging results, as presented in this paper, were obtained.