Structure tensor based in-loop filter for depth video coding

A structure tensor based in-loop filter to remove the compression artefacts in depth video coding is proposed. Instead of involving all neighbouring depth pixels in the joint filtering process, the algorithm chooses one neighbouring depth pixel as the filter output from a reliable neighbour region which is adaptively determined by a structure tensor analysis. It can effectively reduce the negative influence of the non-similar neighbouring depth pixels and improve the quality of the synthesised virtual views. The experimental results show that the proposal is more effective in improving the depth video coding efficiency.