The Len(n)a Dilemma
An appreciable amount of
the research in The Skin Appearance Laboratory is related to problems that arise
from the use of consumer-grade capture and display devices in clinical applications.
The Lena image and the associated concerns over its appropriateness constitute
a microcosm of many of these same issues.
The Lena/Lenna image has become both a de facto standard and an image quality icon. It was scanned at the USC Signal and Image Processing Institute in 1973 from the centerfold (as Lenna Sjooblom) of the November, 1972 issue of Playboy. The SIPI lab at USC needed a set of digitized test images at a time when digitized images were not all that common. An image with skin tones and fine detail met their requirements. The 100 samples per inch resolution of the scanner and a 512 by 512 pixel digital format combined to produce the now classic over-the-shoulder head shot. Perhaps more than any other test image, Lena has been subjected to manipulations affecting the image quality of skin reproduction. As any online image search will demonstrate, there are hundreds and hundreds, if not thousands, of technical papers that uncritically adopt this image as an exemplar of skin color, texture and tone.
Upon closer inspection, the Lena image also serves as a representation of the dangers that aesthetic appeal can overwhelm the proper appreciation of negative technical consequences. Does this image have adequate pixel density to capture the required skin detail when the sensor mosaic is projected onto the surface of the subject’s face? 512 pixels span the image, the pupils are 65 pixels apart and the mean interpupillary distance in adults is 63mm (slightly less for women). These measurements indicate that the Lena image has slightly more than one sample per millimeter (as measured on her face). Is this sampling sufficient to capture skin detail at a level commensurate with that of most current digital SLR cameras? Worse yet, the Lena image is a digital scan of a halftone print and, in consequence, it incurs increased limitations on the representation of fine detail in the published image as compared to that in a continuous tone photographic negative taken of a live subject. The digital sampling of a halftone pattern will result in an appreciable amount of corruption of the high spatial frequencies in addition to the potential for aliasing.
There are also limitations to the use of this image as a functional and accurate (two criteria that together are more restrictive than a metric that only needs to provide simple consistency through repeated use; these two criteria require not only the preservation of the gross features characteristic of skin but also that these details be captured at a fairly high resolution) test of the effects of compression algorithms on images of people, of human skin. The use of halftone prints in magazines is a good example of the pressures mass market items undergo to reduce cost. The result is that these pressures compromise image quality (cf. fine art reproduction of photographic prints). Analogous economic pressures continue today in the design of consumer-grade color cameras and displays that result in corruption of the images, usually aesthetically pleasing corruption, but departures from reality nonetheless.
Well before the digital
era and Photoshop, airbrushing of photographs was used to remove unwanted
imperfections from images. The images presented in Playboy magazine certainly
raised this process to a fine art in the aesthetically pleasing representation
of skin. The Lena image has most certainly been altered in this respect and the
validity of its use (again, apart from consistency) as a test of the
compression of images on actual skin detail is suspect. Airbrushing is the archetype
of utilities now commonly used for digital in-camera signal processing. These
devices regularly perform aesthetic manipulations such as edge sharpening,
color saturation boosting and noise reduction.
The use of airbrushing also represents a caution with respect to the cost of the aesthetic to the functional, and how difficult it is to find fault with an aesthetic icon. The Lena image has been commonly accepted as a standard for the representation of skin, and this use continues today in technical publications of the highest repute. While these compromises might be acceptable in the context of forming aesthetic judgments, their use in the medical domain is far less clear.
Brian C. Madden