第1个回答 2011-06-05
Abstract:
Face recognition is a challenging task, it gathers the multiple disciplines knowledge and technologies, such as signal processing and intelligent control, pattern recognition and machine visual, etc.
Nonnegative matrices decomposition (Non - negative Matrix Factorization Lee him self, the NMF is passed on the Matrix) algorithm is introduced, makes a nonnegative property constraints rebuilding image is not decreased by base images into superposition combination, more accord with human thought "local integral" concept. Better realize the NMF method of face library of local component or components extraction. But and PCA, ICA method is same, NMF methods cannot eliminate illumination, attitude to the influence of factors such as identification. This paper will wavelet decomposition and NMF method unifies, to minimising illumination, attitude to the influence of factors such as identification. Meanwhile, the recognition results and the NMF methods of traditional PCA method recognition results were compared and analyzed the characteristics of the two methods. Points out the NMF method in local feature extraction and recognition has good effect.
第2个回答 2011-06-05
Research of Algorithm for Face Recognition Based on NMF
Abstract: Face recognition is a challenging study which needs knowledge and technology in many subjects, such as signal processing, intelligent controlling, pattern recognition, machine vision and so on.
With introducing non-negative constraints to matrix, NMF(Non-negative Matrix Lee Factorization) makes the reconstructed image constituted by superposition of the basic images, which matchs the concept of parts constitute overall better. NMF method extracted partial face database component or components better. However, the same with PCA and ICA, NMF method can not eliminate the effects of light, gesture and other factors. In this paper, the methods combination of wavelet decomposition and NMF Minimize the impact of light, gesture and other factors to the recognition. At the same time, compared the recognition results of NMF method and traditional PCA method, the paper analyzed the features of these two methods and pointed out that NMF method is better in local feature extraction and recognition.
Key words:Subspace, Face Recognition, NMF
第3个回答 2011-06-05
A Research on NMF_based Algorithm for Human-face Recognition
Abstract,
Human-face recognition is a quite challenging subject, which demands multidisciplinary knowledge and technique such as those of signal processing, of intelligent controlling, of pattern recognition and of robot vision etc..
The Non-negative Matrix Lee Factorization (NMF) algorithm, through imposing non-negative constraints on Matrix(es), re-constructs a human-face image by a series of non-subtractive overlappings of Base Images, thus matching the natural logic in human reasoning--"parts comprise a whole". Comparing with the PCA or ICA method, NMF better extracts parts or components from human-face library, but it cannot eliminate the disturbances from lighting or posture etc. either.
This paper combines wavelet resolution and NMF to reduce to the maximum extent the influences from lighting or posture etc.. It also compares the recognition results of NMF with those of PCA, analyzes the traits of these two methods, and points out that NMF method can better extract and recognize partial or componental features of human face.
Keywords, Subspace, Human-face Recognition, Non-negative Matrix Lee Factorization