Current Proceedings on Technology
Yazarlar: Ceren Güzel, Hasan Sakir Bilge
Konular:-
Anahtar Kelimeler:Face recognition,Dimensionality reduction,Linear discriminant analysis,Block based discrete cosine transformation
Özet: Data in digital world are huge dimensional to analyze and store because of recent technological advances. These large amounts of data include many redundant and irrelevant features. To correctly detect patterns in data, there is a need to remove redundant and irrelevant features from data. In this work, a novel approach that rests on two dimensional Linear Discriminant Analysis (LDA) and block based Discrete Cosine Transformation (DCT) is presented to solve dimensionality problem in face recognition. Using LDA algorithm, discriminative DCT coefficients are extracted from all block based DCT coefficients. To test performance of presented approach on face images, experiments are conducted on Yale and ORL face databases. It is obviously seen that proposed approach reduces dimensionality and accurately classifies samples on these face databases.