LBP vs SVD for Color Images Feature Extraction

Abstract

Images are important element in our daily life. We use images to identify things or associate them with others that become important in decision making or correlate with other object for identification purpose. Feature extraction need to be applied to images for them to be useful for the mentioned purpose. By definition, feature extraction is the transformation of input data into a set of features, normally distinctive properties of input patterns that differentiate between many input of patterns. It is also the process to represent raw image (important criteria or parts of) to help in decision making such as pattern detection, classification, recognition or image processing or machine learning. The process of extracting information from images is one of the important elements in authenticating process of biometric system. Many techniques for extracting features have been emerging in the current research. LBP (Local Binary Pattern) has been a popular technique in biometric systems used for feature extraction. LBP is a texture feature extraction method which produce clearer image result. Another favorite technique known as SVD (Single Value Decomposition) has been selected, discussed and compared with the latter. A sample of still color images were tested for both methods and the results are compared and discussed further in here. This paper concentrates on these two techniques for image processing in a biometric system.