ABSTRACT: Digital halftoning is the technique used by binary display devices to create the illusion of continuous tone. Methods of halftoning, typically, fall into one of two categories: amplitude modulated (AM) or frequency modulated (FM). In AM halftoning, the smallest, indivisible printed dots or pixels are clustered together to form a regular pattern of, sometimes, large macro-dots that vary in size according to tone. FM halftoning disperses the printed dots, usually in a random fashion (blue-noise), with the average spacing between dots, not the size, varying with tone.
The high definition standard of digital printing, FM halftoning leads to spatial resolutions in resulting images much higher than those achieved by AM, but while dispersing dots creates halftones that are least visible to the human eye, printers are rarely capable of printing isolated dots with the consistence required for accurate reproduction of tones. These printers, therefore, resort to AM halftoning algorithms which, by clustering pixels of like color, create halftone patterns that are reliably reproduced, even in unreliable devices. The drawback of this robustness is that by arranging macro-dots in a regular pattern, AM patterns add a periodic texture and, in the case of color printing, suffer from interference patterns known as moire.
Today with the increased interest in digital printing due to the market success of desktop printers, digital copiers, and industrial presses, new alternatives to AM and FM halftoning have become a necessity as printers are achieving higher and higher print resolutions. Isolated dot patterns such as those created by FM halftoning are no longer feasible for many devices while AM halftones do not deliver the image quality of lower resolution FM devices. In order to digital printers to evolve to meet the demands of tomorrow's marketplace, a fundamental change must occur in the halftoning process itself.
In this dissertation, we introduce that fundamental change by studying a new generation of halftoning techniques (hence the title ``Modern Digital Halftoning''), green-noise--AM-FM hybrids composed of randomly distributed and randomly shaped minority pixel clusters that vary in both their spacing between and in their shape according to tone. Being an AM-FM hybrid, green-noise constitutes a tunable model with the ideal cluster size a function of the reliability of the target display device. Green-noise, therefore, bridges the gap such that unreliable devices are relieved of the problems associated with periodic structure and moire while reproducing tone with accurate and consistent results. In addition to describing the statistical properties that define the green-noise model, this dissertation also looks at aspects of green-noise of paramount importance to demonstrating that green-noise is a viable alternative to current techniques. These aspects include computationally efficient halftoning algorithms, adaptive error-diffusion algorithms, and the application of green-noise to color halftoning.
APPEARED: Ph.D.E.E. Dissertation, University of Delaware, Newark, DE. USA, March 1999.
SPONSORS: This work was sponsored, in part, by the National Science Foundation under grant CDA-9703088