Surface color segmentation
AssigneeInterDigital CE Patent Holdings, SAS
InventorsElena Garces, Erik Reinhard
Life of this patent
Estimating the term…
Abstract
In image segmentation it is usually assumed that all the surfaces of a scene are reflecting light in a perfectly diffuse or Lambertian manner. However, most existing surfaces are non-Lambertian surfaces. Examples of non-Lambertian surfaces are specular, glossy or mirror-like surfaces whose appearance may change depending on the point of view of the scene. The method according to an embodiment of the invention does not make the assumption that surfaces of objects in the view are perfect Lambertian reflectors by providing a surface color segmentation which takes into account the change in appearance due to the nature of the surfaces.
Claims20
A computer implemented method for segmenting at least one view of light-field content, comprising:
- selecting a reference view and at least one other view from the light-field content;
- determining a color palette from the reference view, wherein the color palette has a plurality of colors corresponding to a plurality of labels used for segmenting the reference view and the at least one other view;
- associating at least one pixel of the at least one other view to one of the plurality of labels based on a difference between an actual color of the at least one pixel and a color corresponding to the one of the plurality of labels;
- determining, for the reference view and the at least one other view, at least one set of super-rays, wherein a super-ray is defined as a group of rays belonging to a same area of a scene with a uniform color, and wherein each super-ray has a color cohesion measure and a fuzziness measure;
- and
- regrouping groups of pixels whose labels have similar colors in a same set of pixels by determining if a set of the at least one set of super-rays intersects a boundary between the groups of pixels.
A device for segmenting at least one view of light field content, the device comprising at least one processor configured to:
- select a reference view and at least one other view from the light-field content;
- determine a color palette for the reference view, wherein the color palette has a plurality of colors for the reference view corresponding to a plurality of labels used for segmenting the reference view and the at least one other view;
- associate at least one pixel of the at least one other view to one of the plurality of labels based on a difference between an actual color of the at least one pixel and a color corresponding to the one of the plurality of labels;
- determine, for the reference view and the at least one other view, at least one set of super-rays, wherein a super-ray is defined as a group of rays belonging to a same area of a scene with a uniform color, and wherein each super-ray has a color cohesion measure and a fuzziness measure;
- and
- regroup groups of pixels whose labels have similar color in a same set of pixels by determining if a set of the at least one set of super-rays intersects a boundary between the groups of pixels.
Dependencies are inferred from the claim text.