Adaptive DCT based depth map resampling for depth enhanced multiview video coding

In this paper, we propose an adaptive downsampling pattern selection method for depth enhanced multiview video coding taking into account local characteristics of the depth map. The video content is divided into different stripes and a frequency-based measure is deployed to estimate the amount of degradation caused by downsampling and/or coding. The sampling pattern is determined for each vertical/horizontal stripe within a video in such a way that preserves the rectangular shape of the frames, hence the video sequence can be coded with any standard encoder. Simulation results show that the proposed method outperforms the reference by 0.18 dB of Bjontegaard delta Peak Signal-to-Noise Ratio (PSNR) or equivalently brings up to 22.56% and average of 9.84% of Bjontegaard delta bitrate reduction (dBR). It is also shown that the proposed algorithm reduces the complexity of the process to find a proper pattern to about 1% compared to the case where a full search is applied.