Abstract
Shape is considered as a significant characteristic in object detection. In applications such as recognizing the shapes and classification, retrieval of shapes, extraction of shape features and their depiction plays an important role. We propose a shape descriptor which is computationally economical, compact and fast. In this work, we present three methods of feature computation based on chords. Among three, good result is obtained by the third method. Since the proposed work is based on centroid involving a global histogram, shape descriptor is invariant to all transformations. When compared to other shape signatures the proposed shape descriptor resulted in bull’s eye score of 84.56% on MPEG-7 dataset.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have