Perceptually Motivated Shape Context Which Uses Shape Interiors
Source: Nanyang Technological University
In this paper, the authors identify some of the limitations of current-day shape matching techniques. They provide examples of how contour-based shape matching techniques cannot provide a good match for certain visually similar shapes. To overcome this limitation, they propose a perceptually motivated variant of the well-known shape context descriptor. They identify that the interior properties of the shape play an important role in object recognition and develop a descriptor that captures these interior properties. They show that their method can easily be augmented with any other shape matching algorithm. They also show from their experiments that the use of their descriptor can significantly improve the retrieval rates.