Temporal residual data sub-sampling in LDV representation format

The paper presents an analysis on temporal residual data sub-sampling in the layered depth video format (LDV). First, the LDV format with main view and residual views or data is introduced. Then, the extraction of residual data is presented and its block wise alignment for better coding efficiency. Next, the temporal residual data sub-sampling is shown together with an advanced merging method prior to sub-sampling for preserving the necessary information for good view synthesis results. These synthesis results are shown for intermediate views, generated from the uncoded, as well as coded LDV data with different temporal sub-sampling factors and merging methods for the residual data. The results show, that temporal residual data sub-sampling with data merging can outperform regular LDV without sub-sampling for the coded and uncoded versions.