“…Nowadays, many NR metrics are based on the use of neural networks and, synthetically, computed a quality value based on the "experience" made on a set of training images, e.g., [13] . Such methods are trained on commonly used test images not effectively representing the range and variety of different scenes a director or artist may have recorded [31] . Furthermore, the film bases and technologies used during the years, such as Technicolor, Kinemacolor, Tinted and Toned films, present specific features and colours that differ fundamentally from a natural pristine image.…”