Spatial scalable video coding using a combined subband-DCT approach

A combined subband-DCT approach for spatial scalable video coding is presented. The high-resolution input signal is decomposed into four spatial subband signals. The low-frequency subband is used as the low-resolution signal and is separately coded in the base-layer bitstream, and the high-frequency subband signals are coded in the enhancement-layer bitstream. The low-resolution signal is reconstructed from the base-layer bitstream and the high-resolution signal is reconstructed using both the base- and the enhancement-layer bitstream. Similar to MPEG, DCT-based hybrid coding techniques are applied for the coding of the subband signals, but an improved motion-compensated prediction is used for the low-resolution signal. Additionally, SNR scalability is introduced to allow a flexible bit allocation for the base and the enhancement layer. Experimental results at a bit rate of 6 Mbit/s show that the reference coder MPEG spatial scalable profile (SSP) leads to a loss of more than 2.2-dB peak signal-to-noise ratio (PSNR) compared with nonscalable MPEG-2 coding at the same bit rate, whereas the proposed combined subband-DCT scheme is able to achieve a decrease of less than 0.4 dB in PSNR.