FACE RECOGNITION BLOCK BASED STATICAL DCT AND TEXTURE

In this study, a feature extraction methodology proposed for face recognition. The
proposed methodology uses combined block DCT and texture feature. That is, the feature
extraction used is combination of frequency and spatial domains. The frequency domain
feature is statistical block based discrete cosine transformation of a small number of low
frequency coefficients. The texture part of the feature vector used is based on cooccurrence
matrix of the images higher frequencies. The classification vehicle used in the
study is the micro-classifier network. The micro-classifier network is a deterministic four
layers’ neural network, the four layers are: input, micro-classifier, counter, and output.
The network provides confidence factor, as well as proper generalization. Also, the network
allows incremental learning, and more natural. The overall proposed face recognition
methodology was tested using the standard ORL data set. The experimental results of the
methodology showed comparable performance.