Effective method for Shape based Image Retrieval using Pyramid of Histogram of Oriented Gradients

Abstract

Shape Based Image Retrieval (SBIR) is a technique to retrieve images from the database by extracting the shape features of the object. In SBIR, shape features are extracted and then similarity matching is carried out based on some measure. Pyramid of Histogram of Oriented Gradients (PHOG), an improvised version of Histogram of Oriented Gradients (HOG), is very efficient for calculation of feature vector of an image. In this paper, PHOG along with Locality Sensitive Hashing has been exploited as a novel approach to SBIR. Locality Sensitive Hashing (LSH) is used for bit-level similarity matching.