Silhouette Image Classification ssssusing Bag of Local Features

Shape is an important feature to identify an object in the image. Classifying objects by using their shape has been an interesting and important area in computer vision. It has improved a lot in the last decades. In this paper, an approach is proposed for shape classification that uses both local and global image representation using Histogram of Oriented Gradient (HOG), Pyramid of Histogram of Orientation Gradients (PHOG) and CENsus Transform hISTogram(CENTRIST). These features are extracted from MPEG-7 Part-B silhouette dataset and bag of feature approach is used to classify the silhouette image. The performance of proposed method is compared with state-of-the-art methods and analyzed the performance of the descriptors in silhouette image classification. By evaluating these descriptors it can be concluded that the combined HOG, PHOG give better performance than CENTRIST in the context of silhouette image classification.