GUI Based Large Scale Image Search with SIFT Features

Jerrin Varghese

Abstract: Any image consists of descriptive features that uniquely describes the characteristics of the image. The scale invariant feature transform descriptors (SIFT descriptors) are one such descriptors that will not change on resizing of image. The patches generated from scale invariant feature transform descriptors are all converted to unit length. The main concern in searching of image is matching. The visual vocabulary consists of collection of features and this can be clustered using the k means algorithm. The two main k means are approximate k means and hierarchical k means. The clustering helps in indexing of image features. More over the words represented as code words which is in binary form helps in minimizing the memory cost to a great extend.

Keywords: Image Search, GUI, SIFT, K means, Indexing