Coding of depth maps

摘要

Various implementations are described. Several implementations relate to filtering of depth maps. According to a general aspect, a first depth picture is accessed that corresponds to a first video picture. For a given portion of the first depth picture, a co-located video portion of the first video picture is determined. A video motion vector is accessed that indicates motion of the co-located video portion of the first video picture with respect to a second video picture. A second depth picture is accessed that corresponds to the second video picture. A depth portion of the second depth picture is determined, from the given portion of the first depth picture, based on the video motion vector. The given portion of the first depth picture is updated based on the depth portion of the second depth picture.

主权项

1. A method comprising:
accessing a first depth picture that corresponds to a first video picture; determining, for a given portion of the first depth picture, a co-located video portion of the first video picture; accessing a video motion vector indicating motion of the co-located video portion of the first video picture with respect to a second video picture; accessing a second depth picture that corresponds to the second video picture; determining a depth portion of the second depth picture, from the given portion of the first depth picture, based on the video motion vector; and wherein the depth portion of the second depth picture is determined by applying filter weights to the given portion of the first depth picture, and a value of a particular one of the filter weights decreases as a temporal distance between the second picture and the first picture increases; and the value of the particular one of the filter weights increases as the motion indicated by the motion vector decreases.