Integrating Metaheuristics in Computer Vision for Real-World Optimization Problems

7Fusion of Phase and Local Features for CBIR

Pooja Sharma

Department of Computer Science & Engineering, I K Gujral Punjab Technical University, Jalandhar, Punjab, India

Abstract

In this chapter, content-based image retrieval is carried out using global and local features of images. The complete process is based on concepts of amalgamation. The synthesized features of Zernike moments magnitude and phase coefficients are applied to designate the global features of images with local features, which are obtained through histograms by considering distance of centroid of image with linear edges fetched using Hough transform. The processing speed is increased by incorporating efficient algorithms for the calculation of Zernike moments and Hough transform. Overall similarity among training and test images is computed using Bray-Curtis similarity measure, which provides normalized distances and helps upgrade the retrieval rate. The proposed method is capable of retrieving transformed images as well, i.e., the proposed system is robust to photometric and geometric transformations. The results of our comprehensive experiments establish the dominance of the proposed synthesized methodology over other similar methods.

Keywords: Zernike moments, Hough transform, rotation invariance, precision and recall

7.1 Introduction

With the progression in the use of images both in the industry and academic world, researchers are encouraged to develop accurate and efficient methods to retrieve relevant ...