The effect of image rotation on UTV decomposition

Since the singular value decomposition (SVD) consumes high computational complexity on updating its eigenvectors and eigenvalues when new data are included, an alternate rank-revealing orthogonal decomposition that can eliminate this problem such as the UTV decomposition is one of our particular interest. This paper presents a study on directions of principal structures of the images and their effects when the UTV decomposition is employed. The relationship between the UTV decomposition and SVD is also explored. The proposed image denoising algorithm illustrates that the UTV decomposition can efficiently decompose images with vertical/horizontal structures into only a few component as well as the SVD.