Level Sets Based Feature Extraction For Face Recognition

As an important branch of Biological Identification technology, face recognition technology has awide potential application prospect in various monitoring, financial, security and other fields, and is alsoresearch issues in the field of pattern recognition and computer visual field. A typical face recognitionsystem includes five modules: face image acquisition, preprocessing, feature extraction, decisionclassification and result output. Among them, feature extraction is the core problem in the research field offace recognition, it determines efficiency and recognition performance of the subsequent calculation. Atpresent, there exists vector extraction method based on the geometric feature, feature vector extractionmethod based on the statistical and feature vector extraction method based on connection mechanism,which based on the geometry of the feature vector extraction method includes the characteristic analysismethod and cluster analysis method; feature vector extraction method based on statistical includes linearsubspace method and Hidden Markov Model; feature vector extraction method based on connectionmechanism includes support vector machine method, neural network method and elastic graph matchingmethod.The main work of this paper as follows:1．According to computer face recognition technology and face feature extraction method, the level setmethod for feature extraction, and thus for face recognition.2．Preprocess the image using PCA to reduce the dimensionality of the image to filter out highfrequency information, stretch its gray-scale image to integrate it into the same intensity range, reducing thecomplexity of subsequent processing algorithms.3．The black part of the conversion in dimensionality reduction based on the binary image and set the threshold value, transforming the first image grayscale pixel with the transformed image corresponds to thestatistical calculation of its connected component in order to extract the characteristics of face images.4．Tested the ORL and Yale face database using Matlab software, the results showed that face imagefeature extraction based on the level set method and support vector machine classifier for face recognitionperformance is better.