Algebraic feature extraction of image for recognition

摘要

The extraction of image features is one of the fundamental tasks in image recognition. Up until now, there have been several kinds of features to be used for the purpose of image recognition as follows: (1) visual features; (2) statistical features of pixel; (3) transform coefficient features. In addition, there is another kind of feature which the author believes is very useful, i.e. (4) algebraic features which represent intrinsic attributions of an image. Singular Values (SV) of image are this kind of feature. In this paper, we prove that SV feature vector has some important properties of algebraic and geometric invariance, and insensitiveness to noise. These properties are very useful for the description and recognition of images. As an example, SV feature vector is used for the problem of recognizing human facial images. In this paper, using SV feature vector samples of facial images, a normal pattern Bayes classification model based on Sammon's optimal descriminant plane is constructed. The experimental result shows that SV feature vector has good performance of class separation.