A COMPARATIVE REVIEW OF DWT AND SURF FOR RETRIEVING IMAGE DATA ON A LARGE SCALE

Vipin Gopalakrishnan Nair, Nadir N. Charniya

Abstract: The advent of digital era and advancement in the field of imaging technology has increased the amount of images present in the domain to billions. More and more images are created every moment through one source or the other. It is highly inconvenient in such a case to search for a few images of our interest. It is in such a scenario that image search algorithms gain prominence. Research is highly focused on improving the efficiency and ability of algorithms to handle large datasets as well as minimizing the time for searching a query drastically. Bag of Words model is the approach employed by almost every state of the art method. Algorithms work by detecting the most distinguishing feature in an image, representing it in compact form and matching it with features in other images. Accuracy of the result depends on how well the features are detected, extracted and matched. This paper presents two different methods for content based image retrieval: Discrete Wavelet Transform (DWT) and Speeded Up Robust Features (SURF). DWT employs decomposition of an image by wavelets while SURF represents an image in scale space by blurring and using box filters. In this paper we look into detail of how the mechanism of image search is achieved by both the methods.

Keywords: Content Based Image Retrieval, Discrete Wavelet Transform, Wavelet Decomposition, Speeded Up Robust Features (SURF), Hessian Matrix

DOI: https://doi.org/10.15623/ijret.2017.0609020