Conclusion

We presented a proxy-based compression scheme for multiple depth movies of a scene involving dynamic human action in this paper. We used a simple articulated model as the proxy and joint angles as the parameters that approximate the model for a particular frame. Spatial coherence of the structure between views is exploited using the common proxy model and replacing depths with differences with it. Temporal coherence is exploited using incremental encoding of these differences across time. The scheme provides good compression ratios at acceptable quality levels. The proxy-based scheme provides several controls on the amount of data to be sent. This makes it ideal for sending the captured data for applications like 3D teleconferencing.

Figure 8: Compression Ratio and PSNR results for Ballet, Exercise and IndianDance Dataset with block size=25/50/100 and varying joint angle noise levels, $K$=0,5,9,12, plotted against $k$

Figure 9: Compression ratio and PSNR for Doo-Young real dynamic depth movie dataset.