Frequency weighting and selective enhancement for MPEG-4 scalable video coding

In MPEGA scalable video coding, only a small portion of input data is coded in the base layer, and most signal components remain in enhancement layers. In this paper, we propose a new frequency weighting method to send more sensitive frequency coefficients, faithfully with respect to the human visual system (HVS). In order to implement the frequency weighting method by bit-plane coding, we obtain a frequency shift matrix from the HVS-based frequency weighting matrix. We also propose a fast selective enhancement method using coding information, such as motion vectors and residual image blocks. By applying the proposed ideas. we have improved visual quality of reconstructed images. In order to measure subjective image quality appropriately, we define a new error metric, called as the just noticeable difference error (JNDE), based on the Weber's law.