A Multicomponent Image Quality Assessment based on Singular Value Decomposition

Most of the image quality measures rely on the amount of noise added to the image. Determination of both the noise amount and its type is crucial, as some applications may be sensitive to some noise types more than the others. We propose a new multicomponent quality measure that is based on singular value decomposition (SVD), and provide a way to identify some noise types using different components of the measure. The measure is compared to peak signal to noise ratio (PSNR), and is shown to distinguish between noise types the PSNR is not able to distinguish between. Standard images were used in the experiments.