On the use of image quality measures for image restoration

Image quality measurements are valuable tools, crucial for most image processing applications, and used in particular to assess and compare the image restoration (IR) quality. The objective of this work is to investigate the potential of such measures when used as cost functions (integrated in the global criterion) to enhance the restoration performance. In this paper, the proposed approach uses the Structural SIMilarity (SSIM) index measure which is one of the most appropriate measures as it is inspired from the human visual system (HVS) and relatively simple to compute. For the composite criterion optimization, after initializing the algorithm by the alternating direction method of multipliers (ADMM), a gradient descent (GD) technique is used to minimize the global cost function. Finally, simulations are conducted to investigate the contexts in which such quality measures might lead to the desired IR improvement.