Investigation into Intelligent Image Preprocessor Techniques for Artificial Neural Networks

In this thesis we will discuss the process for data preparation of visual or image data ready for use in Artificial Neural Network systems. The thesis will present these concepts, their location in the broader field and the arguments as why certain practices are considered required for these systems; before presenting a number of novel algorithms that are intended as alternatives with desirable properties. These novel algorithms will then be testing in a practical domain (simulating the challenge of face-detection within a scene), followed up by discussions of their successes and failures. The findings presented show that some of the novel algorithms can show statistically significant improvement in accuracy compared to some of the traditional methods used in the field. This thesis concludes with recommendations in which situations the novel algorithms may (if at all) be suitable for use in future designs and potential avenues for further research.