Geek Page - Stochastic Screening
Will this new technology replace halftones in printing?
Since the turn of the century, photographic images have been printed using the same technique. Called halftoning, it converts shades of gray into small black-and-white dots arranged on a grid. Different shades of gray are simulated by adjusting the dot size and the density of the grid. We have become inured to this approximation, perhaps only noticing it in grainy newspaper images.
But a new technique promises to improve images and allow photographs to be printed with more than the usual four colors. This technique - stochastic screening - could be to halftones what HDTV is to traditional TV. Stochastic - from the Greek word stochos, meaning a guess - refers to a probabilistic method for describing how complex processes behave on average. It's a tool often used in nuclear physics and weather forecasting. Stochastic screening works by defining the desired average outcome for a defined area, then relying on the rules of statistics and probability for the precise arrangement of dots. Many of the problems with halftones are eliminated, although some new ones are introduced.
The biggest problem with traditional halftones has always been the presence of repeating light and dark fringes, called moirés, caused by interfering grids of halftone dots. For example, take a photograph of a fabric with a texture similar to that of the halftone pattern. Regions where the two textures align will print lighter than areas where they don't.
Another potential moiré happens with color images. Color printing is made by separating an image into its component colors, then printing each of the separations in sequence. If the grids of the different overlapping color layers interfere, there will be moirés. But back in the '30s, when color printing was the hot new technology, inventor Alexander Murray found that if the grids for the three main ink colors (cyan, magenta, and black - yellow is too weak to matter) are rotated 60 degrees from each other, the moiré disappears. This compromise works well only with three strong colors.
While halftoning uses a straightforward, deterministic translation scheme - each input level of gray matches a corresponding dot size and position - stochastic screening employs mathematical uncertainties. Depending on the type of stochastic screening used, dots of ink can vary in size, frequency, or even shape. What characterizes a stochastic process is that the magnitude of that variable is not set in stone, but left up to a probability.
In a simple frequency-modulated stochastic scheme, for example, a dark gray is represented by a region in which each pixel has a 70 percent probability of being turned on. So for each pixel, a number between one and ten is randomly chosen, and if the result is less than eight, the pixel is turned on. Which pixels are turned on will vary from print run to print run: only the average result is predictable. But this simple method can result in patterns or clumps by chance. So sophisticated - and proprietary -methods have been developed to "filter" the randomness so no such artifacts emerge. Some methods also examine the status of surrounding pixels to determine whether a pixel should be turned on, but this sharply increases the computational complexity of the method.
When they're done right, stochastic images can put their halftone equivalents to shame. Moiré is eliminated, and thanks to the small size of the dots, images look more photographic. Additional inks ("hi-fi color'') can make the colors stunning. Stochastic screening also does not require most users to change their routines. Only the software inside the image setter - the RIP - is different, so most users will notice only slower RIP times and increased image quality. But professional printers will need to be aware of the screening method used so they can adjust their routines accordingly.
If stochastic screening is so great, why isn't it used for all printing? Unfortunately, each troublesome artifact in traditional screening is matched by a similar problem with stochastic screening. The bane of stochastic screening is the tight control needed over the printing process. After an image leaves the pure digital world, it enters the messy world of film processing. Any variation in film processes can transform the final result. Producing tiny dots is the most challenging aspect of printing - the coarser dots of halftones tend to be much more forgiving. The most successful adopters of stochastic screening have been printers who already tightly control the printing process.
Another downfall is that stochastic screening gives flat tints a somewhat grainy appearance because of imperfections in the way dots are transferred to paper. In a halftone, the regular arrangement of the dots ensures that printing imperfections affect each dot in the same way. But a stochastic screen scatters and clumps the dots randomly, so some dots are more affected than others. However, if the printing is done well, these imperfections are minimal.
Stochastic screening also sacrifices the simple, deterministic relationship between dot size and color found in traditional halftones. Just as larger halftone dots lead to darker grays in black-and-white printing, in color printing, larger dots lead to deeper color shades. Because of the deterministic relationship, if you make the dot size this much bigger, you invariably get that much color shift. With stochastic screens, the relationship between dot spacing and color is much more difficult to control.
According to Don Carli, founder of Nima Hunter Inc., and inventor of the term "stochastic screening," techniques exist to solve these problems. Ultimately, he sees screening as a form of binary feng shui, in which, aided by probabilistic analysis, the interaction of each individual dot with its context combines to produce exactly the desired effect. The viability of this approach - in his words, of "jujitsu versus brute force pugilism" - relies on dropping prices and exponential improvements in microprocessors.
As with any new technology, the early hype about stochastic screening - that it would cut press waste and allow lower resolution scans - hasn't completely panned out. Indeed, stochastic techniques may never replace traditional halftones. At the very least, the beautiful work stochastic screening produces will ensure it a solid niche.
