Image Feature Points Extraction And Its Applications

Feature points extraction is one of the key problems in image processing and pattern recognition. Due to the complexity of images and the variety of requirements, the methods of feature points extraction are quite different.The dissertation mainly addresses feature points extraction methods and their applications in several fields. It includes three parts. The first part is feature points extraction in fingerprint matching. The second part is facial feature points extraction. The third part is cephalometric landmarks extraction for cephalometric measurement and analysis.The main contribution of this dissertation is summarized as follows: 1) A singular points based rough matching method is proposed. A novel quad-neighborhood-minutiae-vector(QNMV) is constructed. A new fingerprint matching method based on QNMV is presented thereafter. The experimental results on FVC fingerprint databases show that this algorithm is quite competitive.2) A R-channel of RGB color space based accurate pupil extraction method is presented. Traditional ASM(Active Shape Model) is introduced. After that, a novel RGB channels based ASM method for facial feature points localization is proposed. This method not only uses the extracted pupils to resolve the problem of initialization, but also improves the performance by searching on each channel of RGB space. 3) An edge-tracing based landmarking method is presented. Experimental results demonstrate desirable performance. We also propose a multi-component ASM based cephalometric landmarking method. The simulating results show that this ASM based method is quite useful when used to offer efficient initial position for the landmarks.