Interactive information mobile services

Abstract

Content-based image retrieval has been intensively research upon with techniques ranging from image feature extraction, representation, indexing etc. The Final Year Project report aims to cover the theoretical background, investigates different approaches in content-based image retrieval and performance analysis of image recognition system which targets mobile applications.

Bag-of-Words framework is adapted into the project where each image is modeled as a collection of features plotted over a histogram. Two feature extraction methods were used - dense sampling based and keypoint sampling based approach. Hierarchical k-means approach is used in clustering and scalable vocabulary tree is used in machine learning.From the testing results, Geometric verification on the initial matching results is proposed to improve the performance accuracy. Two methods were implemented, Harris corner detection and Lowe’s keypoint matching. Both methods increased the performance accuracy but at the expense of longer time duration.