Image Compression via Alternating Least Squares

In numerous application areas such as medical imaging and remote sensing, image compression methods play an important role in managing data storage and transmission costs. In this paper, we present an empirical study on a low-rank approximation using the alternating least squares (ALS) method for image compression. For high-dimensional data, we demonstrate the advantage of a matrix decomposition-based implementation that avoids explicit matrix inversion applied in the direct implementation in terms of computational efficiency. The performance of the ALS method is compared with the singular value decomposition in terms of information loss, and compression ratio.