Depth segmentation in multi-view videos

An apparatus comprising:

• a plurality of sensors, wherein the plurality of sensors are arranged to obtain source view images source view depth maps;
• and
• a processor circuit,
• wherein the processor circuit is arranged to obtain a plurality of source view images from the plurality of sensors,
• wherein each of the plurality of source view images has a source view depth map,
• wherein the plurality of source view images and the plurality of source view depth maps are representative of a three dimensional scene;
• segmenting foreground objects in the three dimensional scene from the source view images and/or the plurality of source view depth maps;
• wherein the processor circuit is arranged to generate at least one patches for each of the plurality of source view images,
• wherein each the at least one patches comprise at least one foreground object,
• wherein each patch of the at least one patches corresponds to a foreground object,
• wherein each of the plurality of patches comprises a patch texture image, a patch depth map and a patch transparency map,
• wherein each patch is based on a portion of the source view image,
• wherein the portion of the source view image is smaller than the source view image;
• wherein the processor circuit is arranged to obtain a background model,
• wherein the background model comprises at least one background depth map and background texture data;
• and
• wherein the processor circuit is arranged to generate an atlas,
• wherein the atlas is based on the patch texture images, the patch depth maps, the patch transparency maps and the background model.

A method comprising:

• receiving an atlas, wherein the atlas comprises a plurality of patches and a background model of a three dimensional scene,
• wherein each patch of the plurality of patches corresponds to a foreground object,
• wherein each patch of the plurality of patches comprises a patch texture image, a patch depth map and a patch transparency map,
• wherein the patch texture image, the patch depth map and the patch transparency map are derived from a plurality of source view images and the plurality of source view depth maps,
• wherein each patch of the plurality of patches is based on a portion of the each of the plurality of source view images,
• wherein the portion of the source view image is smaller than the source view image;
• receiving a virtual viewpoint within the three dimensional scene;
• sorting the patches based on the difference between the position of the virtual viewpoint and the position of the foreground objects corresponding to each patch of the plurality of patches;
• and
• rendering the background model and the sorted patches.

A method comprising:

• obtaining a plurality of source view images,
• wherein each of the plurality of source view images has a source view depth map,
• wherein the plurality of source view images and the plurality of source view depth maps are representative of a three dimensional scene,
• wherein the three dimensional scene is obtained from a plurality of sensors;
• segmenting foreground objects in the three dimensional scene from the plurality of source view depth maps;
• generating at least one patches for each of the plurality of source view images,
• wherein each the at least one patches comprise at least one foreground object,
• wherein each patch of the at least one patches corresponds to a foreground object,
• wherein each of the plurality of patches comprises a patch texture image, a patch depth map and a patch transparency map,
• wherein each patch is based on a portion of the each of the plurality of source view images,
• wherein the portion of the source view image is smaller than the source view image;
• obtaining a background model, wherein the background model comprises at least one background depth map and background texture data;
• and
• generating an atlas, wherein the atlas is based on the patch texture images, the patch depth maps, the patch transparency maps and the background model.