Font Size: a A A

Research Of Face Recognition Algorithms Based On Featuere Extracion

Posted on:2014-06-22Degree:MasterType:Thesis
Country:ChinaCandidate:M LinFull Text:PDF
GTID:2308330473953878Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Face recognition is applied to the fields of the security department, video conferencing, identity authentication, digital control and so on. Compared with other biometric identification technology, face recognition technology has a broad application prospects, especially in the aspect of usability, interoperability and so on. face recognition technology has more advantages. Research on face recognition involves pattern recognition, image processing and neural network etc, and has become a research hotspot of pattern recognition, image processing science. Face recognition process is constituted by three parts:image data preprocessing, data feature extraction and classification. This paper focuses on the feature extraction, particularly on solving the problem of face recognition in the illumination variation, expression variation conditions.In the preprocessing stage, the image equalization, the LOG-DCT-based reprocessing and the Gamma correction are the main reprocessing methods for the facial images, which effectively improve image quality, and reduce the computational complexity, so as to improve the convergence speed of the subsequent algorithm.In feature extraction this paper mainly includes the following aspects:it introduces main principle of the Principal Component Analysis (PCA) method, Linear Discriminant Analysis (LDA) method, and Independent Component Analysis (ICA) method of face recognition algorithms, also their main implementation process, and the existing problems of them. Points out the problems of the conventional algorithms and gives solutions. Through analysis of experimental operation our paper compares advantages and disadvantages of the algorithm. Through the experiment, our paper analyses the advantages and disadvantages of these algorithms with comparison. This paper also introduces the method of face recognition based on logarithm discrete cosine transform domain. According to the disadvantages of time complexity greatly in the face recognition process of solving inverse discrete cosine transform, this paper introduces discriminative analysis (DPA) method, so the face recognition process can skip the discrete cosine transform inverse procedure. Experimental results show that, discriminant analysis (DPA) method combined with the method of discrete cosine transform can get the higher recognition rate in the conditions of changing in facial pose and expression or other circumstances. The process skips the discrete cosine transform inversion steps, and reduces the calculation time, with high recognition efficiency, which should be used in face recognition, and will be a more effective method.In the final this paper presents an improved weighted PCA algorithm by using LDA algorithm based on the power of discriminating ability to identify the main ingredients needed to face recognition to improve the recognition rate, the improved algorithm is mainly to solve classification problems of two types.
Keywords/Search Tags:face recognition, Principal Components Analysis, Linear Discriminant Analysis, Independent Component Analysis, Discrete Cosine Transform
PDF Full Text Request
Related items