Chapter 9

Case study

Image clustering

Abstract

The bag-of-words (BoW) model is one of the most popular approaches to image classification and forms an important component of image search systems. The BoW model treats an image’s features as words, and represents the image as a vector of the occurrence counts of image features (words). This chapter discusses the OpenCL implementation of an important component of the BoW model—namely, the histogram builder. We discuss the OpenCL kernel, and study the performance impact of various source code optimizations.

Keywords

image search

feature representation

clustering

histogram

optimization

The bag-of-words (BoW) model is one of the most popular approaches to image classification and forms an important ...

Get Heterogeneous Computing with OpenCL 2.0 now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.