Content-based Image Retrieval in Image World

This thesis presents an interface to investigate the possibilities for a graphic user interface for an image database to be mapped onto an OpenGL surface. The content of the image was used to determine the placement within the interface and thus the images would be clustered according to their similarity. This would allow the user to browse over the images fluidly and interactively and would also provide relevance feedback. The proposed method focuses upon texture classification and combines Fractal Dimension utilising the Improved Box Counting method and Dominant Colour Diamond Segmentation in order to classify the images according to texture and colour. To determine the feasibility of the fractal method, a comparison was made to the Local Binary Pattern algorithm. These methods were subsequently tested in a usability test to ascertain the effectiveness of these methods in real world situations.