Feature Detection using S-Transform

Images are characterized by features. Machines identify and recognize a scene or an image by its features. Edges, objects, and textures are some of the features that distinguish one image from another. There could be many common features in similar images. But, in those commonalities there lies a distinction in terms of features known as subtle features. Numerous algorithms have been reported to extract features from images. Few of them are reliable. Some of them do well under a constrained environment. Many of them fail miserably under low intensity, noise etc. The prominent features are very well identified by many algorithms, whereas the subtle features are often overlooked. In this thesis an attempt has been made to develop an algorithm to extract very subtle features from a given image. A new method has been proposed on the principle of phase congruency to detect features in images. The proposed method uses S-Transform to calculate phase congruency. The proposed method is able to calculate the subtle features even in the very low intensity images. Finally, an application of the proposed method in fingerprint minutiae extraction has also been demonstrated.