Skip to Main Content
A method for singular value feature extraction of a face image is presented, and its model is established, which includes singular value decomposition (SVD), singular value dimension compression (SVDC), singular value vector standardization (SVVS), and singular value vector arrangement (SVVA). SVDC solves the problems that the information of a singular value feature is redundant and calculation data is huge; SVVS solves the problem that a singular value feature has proportion invariance; SVVA solves the problems that face images with the same class possess the same structure features and face images with different classes possess different structure features. Experimental results on the ORL face database demonstrate that singular value features take on separability, stability and independence, and are valid in feature extraction of face images.
Date of Conference: 20-22 Oct. 2004