Singular value decomposition (SVD) is a technique used to represent a matrix as a product of three matrices, with the middle matrix containing diagonal entries called singular values. SVD can be used for data compression, signal processing, and image processing.