Emerging Topics in Computer Vision

5.6. Application to Computer Vision Problems

Thus far, we have demonstrated the capability of the tensor voting framework to group oriented and unoriented tokens into perceptual structures, as well as to remove noise. In practice, however, we rarely have input data in the form we used for the examples of sections 5.3.5 and 5.4.4, except maybe for range or medical data. In this section, the focus is on vision problems, where the input is images, while the desired output varies from application to application, but is some form of scene interpretation in terms of objects and layers.

The missing piece that bridges the gap from the images to the type of input required by the framework is a set of application-specific processes that can generate the ...