Eigenimages

The method of eigenimages is based on principal component analysis (PCA), and it creates a new space to represent a set of images. Advantages of this new space include a significantly reduced dimensionality and an optimal view of the first order data inherent in the data set. This chapter reviews the foundations required of the eigenimage method as well as the generation and use of eigenimages coupled with examples. Eigenimages are created in a process similar to PCA, but with a modification. The traditional PCA process is: warp images to a single grid, compute the covariance matrix of the images, and compute the eigenimages from the covariance matrix and downselect the eigenimages to use. The chapter also reviews face recognition is a famous application for eigenimages. The method of eigenfaces is first order, which means that when an image is compared to the eigenfaces that all features must align.