Nowadays, a common way to identify and classify similar images within large image sets is to implement what are known in the computer vision field as visual searching techniques. The aim of this project was to build a visual search tool using some of these techniques and analyze their performance, using MATLAB.
The image classification process is based on computing an accurate image descriptor for a query image manually selected by the user or automatically selected by the system, which will then be represented in an N-dimensional feature space. The rest of the images in the database are then represented in the same feature space. The distance from each image representation as a point in the feature space to the query image point in the same feature space determines how similar the images are.
I completed this project at the University of Surrey, UK. For more information on the project click on the link to the repository.
