Your country’s flag on a sunny day: Easy enough to say how it should reproduce in a color photograph. But in black-and-white? There is no hard-and-fast rule about the shades of gray most appropriate to a colored image. Over a period of many years beginning in the early forties, that issue was the subject of some fascinating research by men such as L.A. Jones, C.N. Nelson, H.R. Condit and others. The topic was termed “tone reproduction” and was concerned with how scene elements of varying luminance were best reproduced as shades of gray in a black-and-white photograph.
These gentlemen selected 170 outdoor scenes and used a telescopic photometer to meticulously characterize the luminance of various elements in the scenes. They made many photographs of each, varying film, development, and camera settings, and made dozens of (straight) prints with various papers and printing times. The prints were examined in typical room light (about 100 foot-candles) by many observers and rated for “quality.” Each print was also characterized by a tone-reproduction curve that related luminances of its scene elements to the reflection densities at which they were reproduced in those prints. Surprisingly (or perhaps not), the prints ranked highest for quality all had astonishingly similar tone-reproduction curves. That preferred tone-reproduction curve, three versions of it, are reproduced here. The three curves correspond to three values of Dmax, typical of papers of the era. We have included a straight line of unit slope, which was long anticipated to be the preferred curve, as it spaces reflection density in direct proportion to log scene luminance.
The curve has several salient characteristics. It compresses highlight and shadow tones: Observers’ impressions of “best” prints were heavily influenced by the appearance of the middle tones, less so by highlights and shadows. Separation of those middle tones is exaggerated, with curves for the “best” prints always having a slope near 1.15 through the middle grays: Prints with lesser slope were universally rejected as “too flat.” At the same time, those middle grays are significantly lighter than predicted by the unit-contrast straight line forecast: Prints made with darker midtones that approached that line were dismissed as “too dark.”
We must emphasize that the curve family illustrated is for black-and-white reflection prints of outdoor scenes viewed in typical room lighting. Different curves are obtained (some markedly so) for transparencies, motion pictures, and television images, and “best” for each of these depends on the surrounding lighting. Black-and-white reflective prints exhibit a different optimum curve if spot-lit in a darkened room. Further, the preferred curve is slightly different for portraits than it is for the outdoor scenes detailed above.
But bear in mind that this research was being done during the age when black-and-white reigned supreme, and most of it was done outdoors. While print appearance, which is judged subjectively, is characterized objectively by these tone-reproduction curves, the curves themselves can be computed from first principles. This requires knowledge of things such as lens design, camera settings, film-curve shape, development time, printing times, and paper contrast.
Knowledge of the goal—the most preferred tone-reproduction curve— was invaluable to the development of photographic standards characterizing f ilms and papers, as well as to researchers designing new products for photographers. Readers can be certain these authors relied heavily upon it as we designed and evaluated all manner of new black-and-white products during our tenure at Kodak well into the late 1990s.
And it still has a home today. Whether a black-and-white print is created in a traditional darkroom or is the product of a purely digital workflow, this optimum-curve family retains its value as a guide to what are perceived as excellent prints. In the next issue of PT, we will explain how to create a tone-reproduction curve for a digital camera and inkjet printer, and will compare the results to the ideal curve family presented above.