In this project, the focus is on image compression using Singular Value Decomposition (SVD), specifically following the principles of the well-known "Eigenfaces" experiment. The process involves representing a grayscale image M of size m x n as an m x n real matrix. SVD is then applied to obtain U, S, and V matrices. By utilizing the largest k singular values and corresponding singular vectors, a best rank-k approximation to M is achieved.
The project aims to provide insights into the relationship between rank-k approximations and the reconstruction error, demonstrating the effectiveness of image compression using SVD. Additionally, the side-by-side display allows visualizing the impact of different k values on the image quality.