Interactive image segmentation

Abstract

Traditional methods of face detection have been in use for years, with their own advantages and disadvantages. However, with recent technology advancements, these methods are far from sufficient. New methods of face detection include the training of classifiers such as the usage of Haar features to improve the efficiency. As the trend of photo-taking continues, there has been an increasing need for photo-editing software. This project deals with two parts, the detection of faces within an image, as well as the segmentation of this image into two different parts, namely the foreground and the background. The segmentation uses the constrained random walks algorithm, where foreground and background seeds are defined accordingly. The process is then tested using different types of images as well as different test cases to gauge the limitations of the applications.