Embedded human face image coding with set partitioning in hierarchical trees

For human face images, the region of human face (ROHF) is considered to be the most important part, while the background is allowed to have degraded quality because it is considered to be less important. In this paper, a human face detection algorithm is combined with the set partitioning in hierarchical trees (SPIHT) algorithm for wavelet-based image coding. The human face detection algorithm is employed to automatically determine the ROHF in a human face image. The ROHF mask is generated in the wavelet domain. The wavelet coefficients in the ROHF mask of the LL subband are scaled to ensure that they are encoded with a higher priority. Finally the SPIHT algorithm is directly employed to encode the resulting coefficients progressively. Experimental results show that the ROHF exhibits much better quality than that of the background region at any bit rate. The encoded bitstream based on this approach is fully embedded and supports progressive transmission.