How to Allocate Bits To Optimize Photographic Image Quality
Joyce Farrell · Technical programs and proceedings/Technical program and proceedings · 1998
“What is the most efficient1 way to generate grayscale – by increasing pixels/inch or by increasing bits/pixel?”. Simulations, preference judgements and computational image quality metrics all converge to yield the same answer. The most efficient way to generate high-quality photographic images is to increase the number of graylevels/pixel. Consider, for example, two photographic images printed at the same size: A 3 bit/pixel grayscale image is printed at 300 DPI and takes only 0.0983 megabytes of file space. A 1 bit/pixel grayscale is printed at 1200 DPI and takes 0.5243 megabytes of file space. Although the images appear to be the same size, they do not appear to have the same image quality. Both empirical data and image quality metrics predict that the 300 DPI grayscale image will have higher perceived image quality than the 1200 binary image. Clearly, for photographic image quality it is much wiser to dedicate bits to grayscale (bits/pixel) than to DPI (pixels/inch).