Investigating the uniqueness of Singular Values for image recognition
It has been reported that the Singular Values in the S matrix of the Singular Value Decomposition (SVD) is unique. This allows a reduction in the amount of processing needed to recognize the images by identifying the images based on a few unique values instead of comparing all of the image values.
SVD is a method that transforms matrix A into product USVT where U and V transpose as the orthogonal matrices and S is the diagonal matrix and allows the refactoring of a digital image into three matrices. Using the Singular Values of such refactoring allows representation of the image with a smaller set of values to preserve useful features of the original image. The need to find more effective techniques to recognize images using the least resources possible by finding distinguished unique values is always required because of available resource limitations.
In this paper, the uniqueness of the Singular Values is investigated when applied in image recognition. The study includes applying the SVD on gray images, and of those, mainly on two types of images known as identical and distorted images. The distortions on the images include noise, scaling the images by half and rotating the by 45.
Keywords: Singular Value Decomposition (SVD), Singular Values, Euclidean Distance
Download Full-Text
ABOUT THE AUTHORS
Nesreen Nusair
Mohd Belal Al-Zoubi
Nesreen Nusair
Mohd Belal Al-Zoubi