The multi angular descriptor (MAD): a binary and gray images descriptor for shape recognition

Abstract

In this paper, we present the Multi Angular Descriptor (MAD), a new shape descriptor for shape based object recognition and image retrieval. In the binary case, the MAD descriptor captures the angular view to multi resolution rings from each contour point. Placing the rings in different heights enables capturing multi-level global/local features. In gray level, it captures the weighted distribution over relative positions of the shape points to multi resolution rings around the centroid. The multi angular descriptor is robust to noise and small deformations. Flexible parameters makes the MAD descriptor tunable to specific unique characteristics of the different tasks. The extension of the (MAD) descriptor to gray level shapes, can be seen as an extension of a shape context descriptor to be used with low quality gray level images avoiding poor results of the binarization process. Testing the proposed descriptor on the MNIST dataset [16] and a private dataset using two matching techniques gave better results comparing to the Shapes Context and the Histogram of Oriented Gradients (HOG) descriptors.